Benjamin Guler of EvolveLab joins the show to talk about Veras, the AI rendering tool for Revit, Rhino, SketchUp, Vectorworks, and most recently Autodesk Forma that is used for creating stunning architectural diffusion-based visualizations. Ben takes us behind the scenes to give us a glimpse into its unique features, its development process, and the team's ambition to further innovate it.

EPISODE SPONSOR

Chaos

For a limited time, get a 20% discount on Fixed and Floating Annual License of Enscape. Head to https://chaos-enscape.com and use code RES24 at checkout to supercharge your design workflow.

Episode Links:

• Ben on LinkedIn
• Ben on Twitter
• EvolveLab website
• EvolveLab forums
• EvolveLab on LinkedIn
• EvolveLab on Twitter
• EvolveLab on Instagram
• Confluence NYC event details
• Brooks’ Law
• Ben on TRXL 131: ‘It’s Prompting Us’, with Benjamin Guler

Watch this episode on YouTube or Spotify.

-----

The Confluence podcast is a collaboration between TRXL and AVAIL, and is produced by TRXL Media.

• Learn more about them at https://trxl.co/media and https://getavail.com
• Learn more about the Confluence podcast at https://confluence.getavail.com/podcast
• Learn about the Confluence events at https://confluence.getavail.com

Ben, welcome to the Confluence podcast.

2

It's great to have you.

3

And today we're gonna be talking

about Veras and I think a lot of the

4

things that have been going on behind

the scenes leading up to and during

5

development of Veras, maybe you could

just start off and explain what Veras is.

6

And I, you know, I, the stuff

that I see you guys reposting on

7

LinkedIn is absolutely incredible.

8

It has to make you just like grin

with pride that people are creating

9

amazing things with your tools.

10

So tell, tell everybody what it

is, if they haven't heard of it, if

11

they've been living under a rock,

12

uh, and we'll just go from there,

13

Benjamin Guler: For sure.

14

Yeah.

15

So it's gonna be back, uh, or at least

maybe this is the first in this format,

16

uh, but yeah, it's gonna be here.

17

Yeah.

18

So Ves ves is a, I guess render is, is the

way we, you could say, but it's AI render

19

and it uses a diffusion based rendering

to take your, model, your geometry and

20

diffuse it so that it actually uses.

21

Machine learning models to

actually get a rendition for you.

22

And it's, the way it does that

is it, it, it de noises the image

23

and then it adds noise back and it

essentially takes the way, it adds that

24

noise back as per trained material.

25

And it uses a prompt to guide

what you want to see in the image.

26

So you could have something

like, okay, here's a building.

27

Render it during summer or winter,

or, you know, make it a log cabin.

28

So those sorts of like ridiculous, you

know, prompts you could put in there.

29

And you don't have to be as ridiculous,

but it will actually respect that.

30

It will actually try to get,

okay, this part of the building

31

looks like it could be that.

32

So it's starting to form that way

and then it becomes that it becomes

33

a log cabin part of the building.

34

So, um, it's, it's a, so it's an

application that's integrated within

35

SketchUp, Revit, uh, rhino, we

have a beta version of Vectorworks.

36

It also has a standalone and form.

37

So we have all those different, uh,

applications that we're extending

38

that, you know, in the AC space.

39

So,

40

yeah.

41

Evan Troxel: And, and, and I just

want to throw out there like.

42

You

43

don't have to model all that stuff to

44

get that kind of output.

45

Right.

46

And that's, to me, where this really

becomes a decision making or ideation kind

47

of prototyping tool very much, which is

very different from traditional rendering.

48

Right?

49

Rendering was, was always like a

later stage, or it was a visualization

50

technique a lot of times to use for

communication with clients, with

51

pinups in the office, things like that.

52

And now it's kind of shifted

the weight of the potential

53

of a rendering, quote unquote

54

rendering application because

you could feed it a generic piece

55

of geometry and it fills in the

56

details magically.

57

Uh, and just

58

as a personal anecdote, I have a model

of a, a house that I was working on as

59

an architect and was using Veras to.

60

You guys have these

61

amazing little presets, right?

62

Which are like with Snow and

63

Grassy

64

Hills and, and just to

65

kind of show off because not everybody

necessarily knows what to prompt the first

66

time they're gonna prompt this thing.

67

So you give them

68

a little bit of a shortcut.

69

Head start,

70

Benjamin Guler: right.

71

Evan Troxel: I clicked one

of those buttons and I got to

72

see this house in the snow.

73

It happens to be a house in the

mountains, I showed it to my

74

family and they were blown away.

75

Right?

76

Just

77

blown away.

78

And to get back to that, the

79

images that I see you

80

guys sharing on

81

LinkedIn in these various posts,

it's absolutely incredible.

82

So people should definitely check out.

83

I do You guys have a hashtag that

you try to use to kind of group

84

these together on social posting?

85

Benjamin Guler: #VerasAI usually,

but a lot of times I search #Veras

86

'cause you know, that's easier.

87

It's just that one's a very common

one, so you might not get everyth.

88

So like #VerasRender, that's another one.

89

Randall Stevens: it resonates with me.

90

Uh, Ben, I don't know if you know,

but My personal history is I started

91

a company called ArchVision years

ago, which was actually in the

92

services business doing visualization.

93

And what's interesting, I

94

think about what you're tapping into

is it was kind of counterintuitive that

95

a lot of the earliest stage of where

visualization was wanted or needed was

96

actually just to, um, to service the

customer on, for them to go out market,

97

raise money, move a project along.

98

And of course you want that to be

as photo realistic as possible, and

99

that was counterintuitive to the, the

100

design process hasn't

101

even started.

102

So I need, I want Hyperrealism just

to get the project moving along,

103

and then it's gonna go to zero and

then some design process starts that

104

actually, you know, is the real thing.

105

But I, I could imagine that you're, uh,

definitely able to help fill that gap

106

in those earliest stages of, we just

need, the customer needs something to go.

107

You know, we're not designing.

108

We're just

109

trying to sell.

110

Right.

111

Move the

112

project along.

113

Benjamin Guler: Yeah.

114

that, that also reminds me of, or just

back in when I started in this industry

115

too, is, uh, I, I had a course and it

was like just drawing with like prisma

116

color and markers and pen and actually

just getting renditions of that and

117

using all sorts of like different media.

118

And you have one render and it's

like hand done and this is it.

119

So take this render and, and you know,

you'll won't replicate it the same way.

120

And it's just of this perspective.

121

And you can have five different

iterations that have a complete

122

different color palette.

123

It's much more analogous

to that, where it's like.

124

On the ideation side, where

we're like, okay, let's just

125

look at this and explore ideas.

126

And yes, it looks pretty photo realistic

and that is that counterintuitive part

127

where it's like people get kinda set

into like, this is, you know, they get

128

married to, but it's also very powerful.

129

Like it could be that, uh, but yeah,

it's, it's kind of more like that.

130

It's like this kind of expert

painter that learned how to paint

131

really well, how to fine pixel by

pixel, uh, way where, you know, it

132

could produce these, these images.

133

And then, uh, Evan, you were talking

about earlier about, uh, how, uh, you

134

were kind of using it for like a personal

project and just, uh, testing that out.

135

It's funny 'cause um, it bill, uh,

he's in the process of, of, uh,

136

purchasing a house, uh, our CEO and,

uh, he was doing the same thing.

137

It's like, uh, he's just taking

pictures of the house and looking at

138

it like, okay, what I, I want this, uh.

139

I know this is kind of dated and

this, you know, the kitchen looks like

140

this, but what could it look like?

141

So just like, just kind of having, you

know, the whole, you know, the bones

142

are there, the skeleton's there, you

know, just changing the finishes of the

143

building and he was playing with that and

sharing it with our team, and it's pretty

144

fantastic like that you could do that.

145

Uh, and yes, it's not replicated where

it's kind like a traditional renderer

146

where you kind of, okay, you have the

theme model, you set all your views,

147

uh, or your camera and you have your

materials, your lighting and all that.

148

And then you could, you know, make

a video out of it, uh, and, and

149

start to, you know, it's every,

everything is prescriptive, but you

150

have to be that expert and you have

to devote that time to get there.

151

It's, it's not reproducible right now.

152

It could become there soon.

153

We'll see how, how things are working.

154

But, um, right now it's kind of really

powerful for the aviation, uh, stage

155

where you could just explore ideas and

just kinda see what, what could look good

156

and just kinda have that inspiration.

157

It's like your like

Pinterest board, right?

158

Within your actual project, not,

not just kind of going online to

159

find those inspiration images.

160

Like what?

161

Just take those images and put 'em there.

162

So.

163

Evan Troxel: I, I find that is a super

useful way to use these kinds of tools,

164

which is stop hunting for hours online

for something that it represents what

165

you're looking for, and just prompt

what you're looking for and create it.

166

Co-create it with these tools.

167

Uh, so obviously it's using your guidance

to get there and you will get better

168

at prompting it, and it will give you

better results, uh, by doing that.

169

Um, but, but you're, you're

learning a new way of working.

170

And even for like PowerPoint

presentations, instead of going

171

out and looking for imagery

that I, I need for a PowerPoint,

172

I will

173

use Dolly

174

three to do that with prompting

or mid journey, right?

175

Because it gives me something unique.

176

Number one, it gives me something that

really kind of focuses on what I needed

177

to focus on, because it's very difficult

to find stock photography or imagery that

178

may be exactly what you're looking for.

179

It's a fantastic use for that.

180

And now we can do that

with architectural images.

181

As a designer myself, I

also want to throw out that.

182

Like what Randall was talking about

a minute ago has always been a

183

really difficult thing for architects

to deal with on the client side.

184

And it's not getting easier, which is, it

185

looks done, it looks done,

186

done, when you see these images

and clients don't know what it

187

takes to get

188

there.

189

And so

190

reverse engineering something that you

may come up with during this ideation

191

process into actual architecture

is the skill of an architect.

192

And it takes a new conversation

point with a client to get there now.

193

And I think, uh, that, that is something

that we're kind of working our way

194

through on the

195

design side, which is

196

how to talk to people about this stuff.

197

Because like

198

there is your log

199

cabin

200

in the woods.

201

It's Right,

202

there, it looks

203

done, and it's like, nope.

204

Now we actually have to do

everything it takes to turn

205

that into actual architecture.

206

Um, and we've always

207

had this issue on one

208

way or another with,

with rendering, right?

209

It's like, well, it.

210

Like, my wife and I did a project for my

sister-in-Law, and she, like, we had Revit

211

floor plans and she thought it was done.

212

It's like, we haven't even started.

213

And how do you communicate that?

214

Because it

215

was just space layouts.

216

It's

217

just adjacencies.

218

It's,

219

it's like, does this feel right?

220

We're trying to figure out if

we're going in the right direction.

221

And, and they're thinking

because they don't, they're

222

not well versed in this, man.

223

It's, it's done.

224

Right.

225

Can we, let's get

226

started on the project.

227

And it's like, oh my

228

gosh, we have so much work to

229

do and this is

230

that on steroids.

231

And I just

232

wanna throw all that

233

out because these are the kinds of

conversations that architects are

234

having amongst themselves

235

on their teams as they're

pinning stuff up on the wall and

236

sending images out to clients.

237

And, um, I mean, you're, you're

probably getting a lot of heat

238

even back, I would imagine, as

a developer of a tool like this.

239

Like, you're killing our process

that we've gotten so good at.

240

It's com It's changing it, right?

241

You're pulling the rug

out from under our feet.

242

I

243

bet you, you hear all

kinds of weird stuff like

244

Benjamin Guler: Yeah.

245

Now I, I want to, I wanna touch on that.

246

Like, I think it's, I really do

think it's a co-authoring approach,

247

regardless, like you talked about

communication with the client.

248

Yes.

249

That's great.

250

Like as architects and as designers,

you can look at an image of a building

251

and know the project you're working

on and visualize it in your own

252

mind, what you envision that to be.

253

Because you've, you're trained and you've

done it so many times that that's kind of

254

like, you know, you can read a four plan

and you can understand what the layout is

255

with a client.

256

Like it's emotional.

257

Like you can't depict that, you

know, simply by showing a floor

258

plan or like how, how things are.

259

If you actually show a, a imagery of that,

like it's, it's kind of the equivalent of.

260

Teleporting what, how you are thinking

in something that's, you know, something

261

a bit more consumable, something that

could actually be interpreted and,

262

and invoke that emotion with a client.

263

You can say, okay, from these five

very different ideas, it seems like

264

you're gravitating towards this thing.

265

And then, okay, what is that

reverse, reverse engineering process?

266

I think that process actually is also

creative because you're still kind of

267

doing that when you're kind of going

deep into your mind to kind of sculpt

268

out or, or, or re you know, re retract

more things that you've learned over

269

the past and compose this new project.

270

It's kind of similar to that.

271

I think that that kind of synergy is

still there, but at least you are kind

272

of have a much better communication and

more, uh, connectivity with the client.

273

So I think, I think it's a really

kind of, it's not taking away like,

274

okay, like now we don't get to render.

275

Like I think it's a different process.

276

It's actually a hybrid thing

where you're using it even for

277

yourself actually as a designer,

you can use it as an inspiration.

278

Like, oh, I didn't think of, you

know, that kind of materiality here,

279

but okay, how could it really work?

280

And then you kind of design and you

build a building and you design the

281

building in such a way that you.

282

You know, you can

283

put

284

those pieces of the

puzzle together properly,

285

Evan Troxel: Yeah.

286

Benjamin Guler: so

287

Randall Stevens: maybe,

288

uh, maybe Ben, you can take us back

to, uh, just in the development of

289

this, when, uh, what's the backstory?

290

When did, when did

291

it happen?

292

Was there a,

293

when did the technology,

uh, present itself?

294

And then

295

ultimately, how did that evolve

into the product that it is and,

296

and how did y'all think about that?

297

Benjamin Guler: of course.

298

Yeah.

299

So we, I think in 2022, I believe, uh,

it's when we had like merger mid journey

300

V one come out and everyone was super

excited and we saw these kinda like

301

painterly, like ghost-like imagery.

302

It's like V one, V two.

303

And then during that time I

was like very intrigued by it.

304

I'm like, oh, this is interesting.

305

So I made an account and, and started

to play with that and it was awesome.

306

It was very fun.

307

And I was using it for, I

think we were working on.

308

Glyph at the time, or not

glyph another product.

309

And we were trying to come up

with a, come up with a logo.

310

So we're like prompting ideas in there.

311

It's like, okay, what's

the logo gonna be like?

312

So it was a really fun process.

313

And then Stable Diffusion came out,

uh, at, at a similar time with, and

314

it was, you know, lower quality.

315

It was pretty, you know, people

were, but it was exciting 'cause

316

it was open source and now a lot

of people could build on that.

317

That's what we build up,

that's what we built on too.

318

Uh, but back then it was very basic

and it was a lot of different, like

319

very poor quality, uh, imagery.

320

But then they upgraded to 1.4, 1.5,

and then, uh, I think at one point

321

they also added image to image.

322

And then like, okay, this is

interesting image to image.

323

What can you do with image to image?

324

And this is before like, you know,

depths and uh, other kind of models

325

came out, uh, or candy models.

326

And so what we wanted to just play

out was like, okay, what if we

327

get, the idea was just to test.

328

What if we get like, uh,

a Revit image that looks.

329

Close to realistic, like the materials

are better, you make the lighting better.

330

Like what if it's like a close

enough like, like digital image that

331

looks kinda like a rendering that's

digital and just a little bit like

332

let AI change it ever so slightly.

333

What does it look like?

334

Will it look better?

335

Can it fit?

336

You know, correct those edges where

it looks very like digital and just

337

make, add that little dirt just to

make it look like, almost like a photo.

338

And that's what, you know, before

anything, like all the other models

339

came out, we just play with that.

340

We had like really cool results.

341

Like, wow, this is cool.

342

I mean, it still looks

like Rev or whatever.

343

'cause we're testing in Rev first, but

it looks better than Rev a little bit,

344

it was so bare bone, but that was,

you know, so we already started

345

developing like something, some

ideas with that, okay, this is great.

346

And that was like middle of, uh, 2022.

347

Uh, and so, and then, um, control Net

came out and like other models came out.

348

I'm like, okay, let's try

that and let's, let's add add.

349

And we already kinda had the pipeline

for that where we, 'cause we started

350

to tinker with it and we started to

add all, all those, you know, other

351

technologies that's being developed.

352

And so as we were adding that, uh, it

really started to become a product.

353

Okay.

354

Like, actually this is like good,

like we should, like, you know,

355

like, let's see what people think.

356

And then I think at the beginning of, of

January, uh, in, uh, I think it was like,

357

uh, January 2nd or something like that,

uh, we wanted to get it before the, the

358

year ended, but you know, the holidays

and they're like, Ugh, it was so close.

359

But anyways, uh, January of, uh, um,

360

for, for Revit it was very basic.

361

It's just like, uh, you know,

limited resolution and, and, and,

362

you know, uh, but it was really cool.

363

It was like a really cool start.

364

And then as things have developed

and things have, uh, upgraded over

365

time, we just kind of kept on adding

and started to build our like.

366

More of our own, um, even our own models.

367

And we start to train and we, we

started to kind of really finesse

368

the process and, uh, kind of take

more, more own ownership of it.

369

But that's kind of how,

like it's nascency.

370

That's how it started.

371

We just kept on tinkering and like,

okay, let's just see what this is.

372

This is fun.

373

And, and uh, yeah.

374

That's kind of how it, uh, started.

375

Evan Troxel: So Ben, can you show

everybody what this actually looks like?

376

I think we've, we've talked about

it, we've alluded to it, but I

377

think now would make a lot of sense

to just show how it's working.

378

Um, and then we can get into

kind of how you got to this

379

point.

380

Benjamin Guler: Sure.

381

Sounds great.

382

So, um, I have a SketchUp model

I like the nav a lot in SketchUp.

383

So, uh, it's a button on the ribbon

here or on the tool, uh, panel.

384

And so as soon as we click that,

uh, UI opens and then, uh, we

385

could see our, our model in here.

386

Uh, change the resolution.

387

And these are the presets that

you were talking about earlier.

388

Uh, Evan.

389

So we have a few different presets here.

390

So if I click this one, you know,

it, um, it starts off right away.

391

One thing that we've learned with

our Helix product is, um, and

392

that's our kind of interoperability

from Sketch Up to Revit

393

so it brings in, uh, sketch geometry

into Native bim, was just a challenge

394

to get people as few clicks as

possible to get like a success.

395

Just like, okay, get me, like, I wanna

click one button and I get something.

396

It's not the right thing,

but it's something.

397

And so we've learned that and

that's how kind we added this,

398

uh, explore mode where, uh, we've

learned from that application where.

399

Even in this one as well, we've kind

of seen how people would like write

400

different prompts in the beginning.

401

Like they just wouldn't

know what to write.

402

Like I just, I don't know.

403

I'm lost for words.

404

So, you know, the, the rendition

doesn't look great if you

405

don't put anything in in there.

406

You have to kind of

define something in there.

407

Uh, we try to help that a little bit

and it sometimes, you know, some, you

408

could get some, some decent results

now without any prompt, but it's, it

409

really does help guide the rendering

as soon as you add some prompts.

410

So this page is kind of like that, where

it's like, okay, this is that single

411

one button click where you could just,

uh, so I'll just try a few of these.

412

I'll just click them.

413

I'll just go for it.

414

Evan Troxel: I.

415

think it's really important to just

sit on what you just said for a moment.

416

Uh, doing something that makes it as

easy as possible for somebody to see

417

some results that matter to them.

418

It's their model.

419

Right?

420

It's not like these presets are

some model that you're supplying.

421

No, it's applying these presets to

whatever I give the Veras application to,

422

to work with as an underlay basically.

423

And it, it immediately gives me as

the user permission to play with this.

424

And I think that is super important

from a development standpoint to

425

kind of reiterate that message.

426

It's like, what do you do when

you don't know what to do?

427

Oh, there's these buttons right here.

428

And that's a great place to start.

429

And, and just for people who are

developing apps, I, I, I see this

430

come up over and over and over again.

431

Like, what would it look

like if it was easy?

432

Because we've all been

through on this call, right?

433

What does the

434

vray interface look

435

like?

436

It's

437

like, are you serious?

438

It's like flying a jet, right?

439

It's, it's, it's got the, the dialogue

boxes and the settings and the, the

440

sliders and the words I've never

seen before as a first time user.

441

And it can't, we, we know what it

takes to be a, an expert V-Ray user

442

with the materials and the lighting

and all the settings, and then there's

443

something like this, which is, this

is a completely different paradigm

444

Randall Stevens: Yeah,

445

Evan Troxel: you're making

it as easy as possible to

446

get started with

447

that.

448

Randall Stevens: we talk about

449

that as time to value, Right.

450

What's your time to value for

451

that customer so they get their

452

kind of taste of success

as quickly as possible.

453

Uh, just a quick backstory.

454

I used to, like in the early, early days,

early two thousands, I was teaching at the

455

university and I used to teach 3D studio.

456

And it was like, I don't know if it was

me or the students, but it seemed like

457

every year that I taught it, it got hard.

458

It took longer and longer to

get them to the point where they

459

could actually just do anything.

460

And then SketchUp came around

and I'm like, man, the time

461

to value is just like nothing.

462

They can start,

463

you know, in, in 10 minutes,

464

start seeing some success.

465

So I agree, and, and I think you

know, what you're doing here,

466

Ben, is, is, is a key to success.

467

Show some instant value and then.

468

They'll give you more time, right.

469

Evan Troxel: these are just amazing.

470

Like, let's just, they're magic.

471

These are just magic.

472

Like I, you're smiling, I'm, I know

like you feel this feeling all the

473

time, but, but it's literally magic when

people see these for the first time,

474

no matter what you feel about AI and

the training and who's getting paid and

475

who, like that, that's, we've, we've

476

talked about those things on this

477

podcast, right?

478

The

479

governance, the ethics, uh, there's

so many things around this, and

480

this is, this is seriously magical.

481

And I just wanted to also point that

out, that these images that you're,

482

that, that your tool is presenting

us with immediately give us that

483

emotional connection potential.

484

To, to respond in real time, and I think

that is an extremely valuable tool to have

485

as a designer.

486

Benjamin Guler: Totally agreed.

487

And it is, it is very magical.

488

Like it never ceases to amaze

me like, oh, that looks cool.

489

The other one also looks cool.

490

So that's kind of part of it,

uh, part of the process of,

491

of, uh, developing the tools.

492

So like fun and so, uh, uh, you know.

493

There's so much fulfillment to

just kind of like, play with it.

494

And so like we just

keep on building on it.

495

Um, one thing that you said earlier

too that I wanna also kind of come

496

back just slightly to is, uh, as you

know with the presets, once you kind

497

of have those presets on here, you

could go back to any one of these kind

498

of renditions and you could see kind

of all the settings that were used.

499

So if you

500

change your, your image here, it updates

the prompt that was used for that.

501

And then you can kind of see,

okay, this is how that was written.

502

Uh, and like I think us as humans

is, we're pretty good at like reverse

503

engineering, uh, or not just reverse

engineer, but if learning by mimicry.

504

So like if you see like, okay, it

looks like those ideas are combat

505

separated and those phrases are built

that way for each comma, you kind

506

of start to read that without even

needing a, you know, a tutorial.

507

Like, oh, okay.

508

That's how that that prompt was.

509

'cause you could write anything,

you might write a narrative

510

and, and con a conversation.

511

Like, but it's not, you could see

the examples, it doesn't do that.

512

So you could, we're pretty intuitive

creatures where you could like, okay,

513

I could see how it's being written.

514

So I think that was kind of part of that.

515

I think you're also talking about

Randall water, like we want to get

516

to that aha moment where it's like.

517

The first click and then it's not

the right image yet, but you got

518

something with your data, like

it's your project and it looks

519

interesting.

520

Okay, now let's make it look like

what you want it to look like.

521

And you've seen an example of, you know,

of what that input, what it produces.

522

So if you wanted to change this,

you know, like let's say I don't

523

want a timber building, I want a

monolithic concrete building, right?

524

Then I'll render a few of those.

525

Then I, 'cause I know that's my

project, I'll just change the parts.

526

Like if you have something you we're

good at, like picking up certain parts

527

that are incorrect or what you want.

528

And so you just change the part

that you want to, what you think

529

it should be for your context.

530

And then, yeah, that, so that was

also part of that, uh, part of that

531

kind of thinking how we, we built like

those two kind of menus where it's

532

just like a separate menu altogether.

533

Randall Stevens: yeah, I think,

534

Benjamin Guler: you go.

535

We have it.

536

Model concrete,

537

Randall Stevens: I think people will

give you infinite amount of time if

538

they got that first hit of value.

539

Because then they will they

will stay in here forever,

540

maybe longer than they should.

541

Right.

542

Um,

543

Evan Troxel: I was gonna

544

say, I was gonna say dopamine, uh,

uh, you said hit of value, but hit

545

of dopamine is what I was thinking.

546

It's just as, as somebody who is

literally addicted to creating images

547

of architecture.

548

Right.

549

Uh, this is, this is feeding that, that

addiction in a, in a very interesting way.

550

Benjamin Guler: right.

551

So I'll, I'll just, uh, let

me reframe this a little bit.

552

Just gonna move it ever so slightly this

553

Randall Stevens: very cool.

554

Benjamin Guler: you go.

555

Yeah.

556

I'm, I'm giggling out here

even just by, by this project.

557

Like I, I, I'm lucky the concrete look.

558

Let's get a few more of 'em.

559

Evan Troxel: So, yeah, this is,

this is what it's like being on

560

a Zoom call with you, Ben, right?

561

Like we're, we're, we're

supposed to be talking

562

about this, but you're

doing that on the side.

563

You're, you're not distracted by

564

Benjamin Guler: a little bit.

565

Evan Troxel: You, you're

not distracted by email.

566

You're, you're

567

actually doing

568

rendering.

569

Benjamin Guler: Yeah.

570

Yeah.

571

Randall Stevens: So

572

what, uh, Ben, can you, uh.

573

Can, like how many images are

people creating when they're like

574

working on a project like this?

575

Do you have any stats on like they

create 10 variations or a thousand

576

variations, or what does that look like?

577

Benjamin Guler: Yeah.

578

So, uh, we do have optional

analytics, uh, that, that we

579

do collect, that kind of stuff.

580

Uh, we don't collect the images

or we don't train on the images.

581

That's a lot.

582

Those questions come up a lot.

583

Sorry, I'm gonna get to the answer.

584

Uh, but in the same vein, like these are

just things in my memory that come up.

585

Um, but

586

like what we've observed is like,

people like to stay within like one

587

shot and just explore with that,

like a hundred images or 200 images.

588

Uh, it was more, more images kind

of in the very beginning because the

589

quality wasn't that great and every

like 20th one was a really good one.

590

So then you're like just going there.

591

'cause like I want that good one.

592

But now almost every rendition

is, is pretty decent.

593

So that number kind of came down a bit.

594

Um.

595

But, uh, yeah, like we have a few

different, like we have our forums and

596

we see some people how their process is.

597

Some people render like the scene

without any objects 'cause they want

598

like a baseline and then they wanna

build on that, okay, I just want

599

my ceiling and like, no furniture

for an interior scene, let's say.

600

And then you build on that your furniture,

you, you build other the pieces and then

601

you have control of all the layers if

you wanted to, uh, to kind of just keep

602

different parts and different pieces.

603

And that's actually we, what we

already offer already inside the

604

application, but it's kind of limited

right now, which we're expanding to.

605

But yeah, to answer the, the original

question, it's, it's, uh, quite a

606

few different images and sometimes

changing on and off, uh, different

607

kind of aspects of the, of the, of the

image and using the seed, like which

608

the seed is, uh, uh, lets you kind of

keep the same, uh, if you were to, it's

609

prescriptive, uh, it's deterministic.

610

So if you were to run into the same

scene again and you turn the seed on,

611

then it will produce the same results.

612

Uh.

613

Uh, so then if you make minor

changes, let's say you remove a door,

614

everything else will stay the same,

but you'll have that rendition of

615

the version with that door removed.

616

So if I wanted to do, for this one,

for example, uh, and I hide, I don't

617

know, part of this, uh, you could, you

could see how, uh, that will change.

618

Let's see.

619

I want to kind of hide a

significant part of the building.

620

Let's see if this could work.

621

Okay, that's probably good enough.

622

So I'll, I'll, I'll leave the same, uh,

um, seat here and then I'll, I'll, I'll

623

render it one more time without, uh, okay.

624

So I'll use this one actually.

625

Okay.

626

There you go.

627

So, I mean, it's gonna take

a second to render, but yeah.

628

So the, the original question, like,

I think people like explore and try

629

different, uh, um, kind of renditions to

get kind of that design and then they kind

630

of finesse, uh, in like, let's say you

like most of the image, then you render

631

more different parts of that, that image.

632

So.

633

Yeah,

634

Evan Troxel: So when it came to you

designing this tool, this application,

635

uh, as, and it, it, it looks the

same no matter what tool you're

636

using as kind of the host, right?

637

It's like you see the same

presets on the, on the side.

638

You see the same row of

imagery across the bottom.

639

Maybe take us through the layout

of this and, and why you decided to

640

do it the way that you did as, as

641

you developed the app.

642

Benjamin Guler: sure.

643

So a bit of a background.

644

My, my background is

also from architecture.

645

I went to school for that and

um, I actually schedule was the

646

first application I learned.

647

Uh, then I learned Rhino and

then, uh, when I got hired to

648

work in the firm, I learned Revit.

649

So, uh, that's kinda was my journey.

650

Uh, but throughout that, even in school,

like I was, uh, always, uh, excited

651

about renditions and like I always

got a better CPU to be able to render.

652

Uh, you know, I.

653

High resolutions just for presentations.

654

So that was always a, a thing

that I was really drawn to.

655

Uh, and so as a, in my, uh, career, uh,

when I was kind of more in that architect

656

role, I would, uh, you know, produce

a lot of visualizations and do that.

657

So I've used tools like Lightroom,

I've used Photoshop a lot, and all

658

those elements kind of, I think

affected kind of the layout of this.

659

So if you look at the bottom

RI room, that's totally,

660

uh, affected by light room.

661

Like light room,

662

you can see it.

663

Yeah.

664

Evan Troxel: Yep.

665

Benjamin Guler: So, uh,

that, that's how that works.

666

Uh, the Photoshop side is kind of

something that we're actually bringing

667

in, but we have kind of, part of it

is we're about to release this week

668

is just the layer system where you

want to build your image and you kind

669

of like, you know, this part of the

image and not this part of the image.

670

So, for example, we

could do that even now.

671

This is a, uh, let's say if we added, uh,

let's say, I'm just gonna make this part

672

something ridiculous, like some kind of

maybe red material for the sake of, it

673

might just call it red or something red.

674

Okay.

675

And you'll see that, um, you know,

you could, you layer your composition

676

when you get to that design.

677

So those kind of elements and those

tools that I've used in the past, uh,

678

and other people in our, in our, in our

team used in the past kind of start to

679

find their, their way into, into how the

application, the interface is, is built.

680

Um, so yeah, so if I go over it a little

bit, we have these three tabs at the top.

681

The first one's kind of like that

onboard, like almost, uh, beginner,

682

like you're, you're learning it.

683

But we didn't want to just keep it for

the, for the beginner because like I

684

go back to it, like when I wanna try

something very, like a quick start.

685

So we didn't wanna make it just, you

know, wanted to have more use out of it.

686

So it's very simple.

687

You're not inundated with any, you

know, any dials or any, any sliders.

688

Then compose the ideas that, okay, you

are taking kind of ownership of this,

689

this, this design and you know what you

want, you know, the materials you want

690

so you can start to explore a bit more.

691

Um.

692

Uh, more restricted, you kind of with

certain materiality and then refine

693

is, okay, I'm set with this view.

694

I like most of it, but I

wanna start to change it.

695

So this is where you can't

change the resolution anymore.

696

You're actually, this is like your

Photoshop mode where you're getting

697

in there and you're kind of making

slide changes in here and everything

698

that we're adding to this kind of tab.

699

And there's a lot of things that

are, is coming, is kind of geared

700

towards that, where it's helping

with, with getting that design, uh,

701

much closer to what your vision is.

702

So, and then at the bottom

703

Evan Troxel: the

704

Benjamin Guler: have,

705

yeah,

706

Evan Troxel: oh, just just before you

go on here, I think one of the things

707

that, that really stands out to me is

that this tool is based on workflows.

708

Of an architect.

709

Right?

710

And, and one of the things that I think

we've all struggled with in the past with

711

Renderers is it produces one image or, or

an animation at a time, and then it's kind

712

of left up to you and your folder layout,

713

folder structure on your

714

drive

715

to kind of flip, flip between images.

716

And that's where Lightroom or Photoshop

kind of changed the game, right?

717

Because they're really workflow based

tools and, and rendering is a workflow,

718

so it gives you the tools you need

when you're in that mode of working.

719

I think that makes a lot of sense to

kind of think about the end user as

720

a developer and what they're doing

in this process and what tools they

721

need at their fingertips right then.

722

And, and that you're, you're pulling that

off here because when I'm ideating, I

723

want to look at all the

724

snapshots and I want to

725

flip back and forth and see what I like

about this one, what I don't like about

726

this one, because that's gonna help me

craft the next prompt for the next one.

727

Potentially or refine an area.

728

And all of those have to do with

my workflow, uh, because that's

729

the mode I'm in when I'm doing

730

this kind of thing.

731

Benjamin Guler: exactly.

732

That's you could, you

said it better than me,

733

so, yeah,

734

Evan Troxel: This episode is made

possible with support from Chaos-Enscape.

735

This is the perfect time

to set good intentions and

736

resolutions for the months ahead.

737

Whether you aim to solve your design

challenges faster, run your projects

738

more smoothly, make quicker decisions,

or share immersive walkthroughs with

739

the click of a button, here's some good

news: you won't need any resolutions.

740

Chaos Enscape has the

best 3d workflow solution.

741

Chaos Enscape is the industry leading

real-time visualization plugin that is

742

100% fast, 100% easy, and fully integrated

into your favorite design tools.

743

It is the trusted choice of over 500,000

monthly users across 150 countries.

744

Starting today, you can get a 20% discount

on fixed and floating annual licenses.

745

Just head to Chaos-Enscape.com

and use code RES24 at checkout.

746

That's Chaos-Enscape.com using

code RES24 at checkout to

747

supercharge your design workflow.

748

Thanks to Chaos Enscape so much

for their support of this episode.

749

And now let's get back

to the conversation.

750

Benjamin Guler: yeah.

751

You start like lower, like big, broad, uh.

752

A big, large gamut.

753

It's like, I don't know,

754

like, I wanna try 10 things and, uh, and

then, okay, I like one of those 10 things.

755

Okay, now let's try, you know,

10 variation variations of that.

756

Okay, I like this variation.

757

Let's make this the real thing.

758

So that's kind of the

Exactly, that's the process.

759

Evan Troxel: It's just like architecture.

760

You start way back and then you zoom

in, and then you zoom out, and then

761

you zoom in and then at some point you

decide, okay, this is the direction

762

and we're gonna zoom in more and more

and refine and refine and refine.

763

And it makes a lot of sense for the

actual user who's using this application,

764

I think, because it, it feels pretty

765

natural then at that point.

766

Benjamin Guler: It's also

like diffusion the noises.

767

And the noises.

768

Evan Troxel: Hmm.

769

Benjamin Guler: I had to do that.

770

Add that in there.

771

It gets clear, the image gets

clear and clear and clear, uh,

772

to, to what it needs to be.

773

So, yeah.

774

Mm-hmm.

775

Randall Stevens: So the, been the, uh,

this ability to mask and layer and build

776

that up, uh, it sounds like that's close.

777

Uh, was that, uh, driven

by a user request?

778

Uh, how, how did that feature,

how did y'all decide that that

779

should be what, what you do next?

780

Benjamin Guler: Yeah, there's quite a

few different examples of how, uh, even

781

personal examples, like I would take from

five images and I would go and Photoshop

782

and, and you know, get elements from

each one that's coming, essentially.

783

So you just don't have to leave

the application to get that.

784

Let's say you have five or these

three images and you like part of

785

one, but not part of the other one.

786

I want to just blend the two and I

want to keep the parts that I like.

787

And then in Photoshop, you know,

it's kind of super powerful.

788

You can do anything you want, but

in Photoshop you have to know like,

789

okay, where you add little mask,

the, you know, the feather of the

790

mask and where, where it is in the,

in the channels, the alpha channel.

791

So you have to do all that, and you have

to kind of know how those things are.

792

And our interface's

just, you know, simple.

793

It's like, okay, you know what you want?

794

You want to combine these two

images, there's just two buttons

795

and you don't have to learn that.

796

Um, so we've seen that actually

with a lot of our end users,

797

users already doing that.

798

Like they're, they're taking their

images and they're, they're going to,

799

to Photoshop it and doing that, so.

800

Okay.

801

It makes sense.

802

Okay.

803

It seems like this is something that, that

you're doing, let's just add it in there.

804

Uh, so, and also personal,

personal experience too.

805

Like, I, I think I've done that, uh,

quite a few times and I think there's

806

a third one, but I'm, I'm forgetting

there's a third example on that.

807

Randall Stevens: Are are, are you

able, uh, and I'm not as familiar with

808

this, with this tech bin, but are you

able to separate material choices from

809

lighting and say, I want that material,

but it needs to now be consistent?

810

'cause that, you know, Photoshop,

that was always the challenge was,

811

you know, you could end up with mixed

lighting and, you know, yeah, you could

812

collage stuff together, but, um, are

you able to control those separately

813

and then use those as separate

passes or separate inputs, I guess.

814

Benjamin Guler: Yes, we could, we

could do that in the live version

815

that we have right now, we don't have

that, but, uh, not so much on the

816

lighting side, but the material side.

817

Uh, definitely on the material side,

the different maps and channels that

818

you could have for all the different

materials that are present in the

819

scene, that's, uh, something that,

uh, is actually, uh, very, uh, man, we

820

could manipulate that a lot actually.

821

And that's very powerful.

822

Randall Stevens: Yeah, I was just

noticing even as you flip through

823

those, the, you know, the lighting,

the where shadows were were different.

824

So if, if all of a sudden in one part of

the image you said, I really like this, I.

825

But maybe what you really like is

the material choices or the way that

826

that was affected.

827

Can you, can you end up re basically

re rendering, holding the material

828

there and changing it over here with

consistent lighting and those things.

829

It

830

seemed like natural

831

evolution of this would be that

832

kind of granularity.

833

I.

834

Benjamin Guler: Yeah.

835

So with, with the lighting, if

you wanted, like if it's uh, you

836

know, overcast, that's kind of

nice 'cause then it just blends in.

837

So if you're kind of have that

part of your prompt, then you don't

838

have to worry so much about it.

839

But if it's direct lighting, then

that's also something that, something

840

that we've seen other users do

because of the, uh, the base.

841

You have actually have the shadows

that had shadows and you know,

842

you have to include in the prompt

what that means, what those,

843

you know, what is that geometry?

844

Is that the carpet changing its color or

is it actually a shadow of the carpet?

845

So you kinda have to add them the prompt.

846

That's what we've seen.

847

And per that, you could then, you know.

848

Allow the engine to, okay, understand

it's the same material, but you're

849

kind of rendering the shadow.

850

So then you can have consistent views.

851

Let's say if you have like a sun

through a window, uh, coming through

852

and you, you, you want it to render

just different materials, but you still

853

want the same shadow in the same spot

that we, we've seen success in that.

854

So

855

Evan Troxel: One of the, one of the

benefits that you mentioned there

856

is just not having to learn another

tool or know another tool to the

857

level of expertise that you needed.

858

Are there additional benefits to you

doing all of this in Veras, like giving

859

those abilities to a user, like like

blending, like you mentioned, feathering

860

and masking and all those things.

861

Are those things being taken care of

then by Veras at a, at a, at a pretty

862

nice level to So where I, I feel like I'm

gonna be pretty happy with the outcome

863

without knowing the depths of Photoshop

to get those kinds of results if I

864

had to do it manually.

865

Benjamin Guler: yeah, I actually

changed, uh, the dev version here.

866

Uh, but that's exactly what

we're working on, uh, this week.

867

I think I could show you a GIF though.

868

It's really cool.

869

I wonder if we, we recorded that gif.

870

Uh, let's see.

871

But yes, yes.

872

The answer is yes.

873

Uh, I don't know if I

have a visual for it yet.

874

Again, this will be probably released

by the time, uh, this, this airs.

875

Uh, let's see.

876

I think I have maybe this one.

877

Okay.

878

I'll show you this one.

879

It's a bit, uh, early, so it's

a, you have to forgive it.

880

cause I think it's really fun to,

to look at and it's really, uh,

881

interesting how it, how it blends things.

882

And that's exactly, uh, what you want.

883

Let's say you want to, um, talking

about lighting, let's say you want it

884

to just change the sky to something

that's very different the way you

885

would want to blend that matters.

886

Like, you really wanna,

887

you don't wanna,

888

Evan Troxel: the, the, building, right?

889

It's not just the sky,

it's, it's the lighting,

890

it's the shadows, it's everything.

891

Benjamin Guler: Right.

892

So here I will do something a little

bit scary, but I think it's gonna work.

893

So this is gonna get reset

for a second here, but I will

894

share my screen in a second.

895

It's pretty neat.

896

Okay.

897

Here, sharing screen and

let's see here, our screen.

898

Alright, cool.

899

So I'm going to just render this

once and then I'll go to the,

900

uh, the mask, uh, to, to show you

guys, uh, how, how that will work.

901

Again, this is without the

layer system, uh, built in yet.

902

So it's, it's all the backend is

already, it's just the UI is not ready.

903

So it's, that's kind of

the, the cadence there.

904

Uh, so let's see here.

905

Uh, this will take a second here.

906

Okay, so let's say that, uh,

okay, let's say I wanted to change

907

the, uh, the sky, let's say.

908

So if we added this and we just say.

909

I'll kind of eat into it, and

you could actually kind of draw

910

here if you drag and click.

911

So I'll just leave it there.

912

And that's a pretty, you

know, funky mask there.

913

Let's just say, uh, night sky.

914

All right, let's see what we get here.

915

And I'll render that.

916

And right now I didn't enable that, that

mask to be kind of feathered properly.

917

So you'll see, you'll see what

I mean, uh, when it, when it

918

renders, uh, what the issue is.

919

But we have this kind of currently

in the, okay, there it is.

920

So pretty, pretty rough, uh, there.

921

But we have this, uh, kind of

feather where we could start

922

to kind of feather that in.

923

You can see here, like I

could go all the way to

924

a hundred, but again, it's, it's early on.

925

So this is getting there.

926

But you can see here like

how, how you want it to.

927

Just further that ever so slightly

that it looks like part of, so if I

928

go back and forth, it looks a bit more

929

correct.

930

Randall Stevens: Makes

931

sense.

932

Benjamin Guler: yeah.

933

We could go here and actually

keep a constant thing.

934

Let's see here.

935

So if I do this, say just red

or red, seems to give some

936

interesting results sometimes.

937

Like red architecture.

938

Yeah, it's pretty decent.

939

Right now we're like around 15 seconds

per render and at 10 24 by 10 24.

940

And we're looking at, um,

getting that even, even faster.

941

A lot faster.

942

It looks like I didn't do it, but

um, so if I look at this one, uh,

943

actually, let me go back to this one.

944

Yeah.

945

So if I want to have a hard edge and

then if I go all the way down, you

946

could see that that's kind of exactly

what that selection was versus kind

947

of blend, blending it a little bit.

948

Cool.

949

Yeah, that's a lot of fun.

950

Having fun all the time, like

playing with it, trying different.

951

Uh, prompts and, and, and,

uh, uh, exploring different

952

projects, interiors, exteriors.

953

Evan Troxel: I've seen

some incredible interior.

954

Renderings.

955

Uh, I think you just posted on LinkedIn,

you reposted somebody's post a day or

956

two ago where they had never used a tool

like this and they were just playing

957

around with, in some interiors of, I

958

think it was like a living room.

959

And it was like, and they, they were

just completely blown away by what they

960

were able to co-produce with a tool like

this in such a short amount of time.

961

And I think that's what's so exciting

about this for people who try it out,

962

who are willing to put in a little

bit, I mean, it's literally like the

963

least amount of effort you could do

964

to get renderings right?

965

and

966

and decent really,

really decent renderings.

967

And, And, then therefore a direction,

uh, you know, a pot, a potential

968

emotional response, which then says,

yes, that's the right direction.

969

Let's go that direction.

970

It's pretty

971

incredible.

972

Benjamin Guler: Yeah, so I'm playing

just with this kind of cabin interior

973

here, uh, chair, you know, wood

ceiling, but I'm using a prompt

974

that's probably not the right one.

975

So let's say we started to change

it to something that's more.

976

Also, we have this, um, I'll

change the prompt soon, but we

977

have this kind of slider here.

978

So if you really want it to stay closer

to the geometry that you have in your, uh,

979

model, you have this geometry of a ride.

980

If you go all the way out, it

will really start to, you know,

981

introduce other elements and go

a little bit more exploratory.

982

Uh, and then we also have the material

override, which tries to keep, you

983

know, the, when it sketch up here tries

to keep those sketch up materials.

984

Um, but yeah, let's, let's say if we,

the issue with the, with the material

985

override is just, it doesn't, um, I.

986

The ketchup materials are not that great.

987

So the more it blends with that, the,

you know, but we're working on that,

988

the, the less interesting it looks.

989

Um, but yeah, let's say if I made this

one, uh, let's say, uh, wood walls,

990

wood walls, uh, wood ceiling, and then

we'll leave everything else to save.

991

Um, but yeah, I think, I think

that's one, one thing that we've

992

seen a lot too with frontiers.

993

'cause interiors are usually hard,

especially in traditional rendering

994

because you have global illumination

that affects so many different services

995

and the way blends, like, were really

good to pick up those inaccuracies.

996

Uh, whereas exteriors is a lot of times

easier because it's like, you know, it's

997

strong lit, you know, or, or hard lighting

from sun, or even if it's soft lighting,

998

like just the way you, you know, in

traditional rendering, it's a bit easier

999

to get to, to, to a good result there.

00:42:11,134 --> 00:42:15,274

Whereas interiors have all these, you

know, subtle reflections and so many

00:42:15,274 --> 00:42:17,404

different, uh, um, material interactions.

00:42:17,704 --> 00:42:22,744

That start to, uh, you know, in

traditional method is, is, is challenging.

00:42:22,774 --> 00:42:25,925

You have to be really good at

and, and learn that process and,

00:42:25,954 --> 00:42:29,164

and, and craft that art, uh, to be

produce, produce decent results.

00:42:29,782 --> 00:42:32,992

Evan Troxel: What's the most kind of

valuable feedback if you had to, you

00:42:32,992 --> 00:42:37,282

know, pick one or a top three of from

users as you've gone through this process?

00:42:37,282 --> 00:42:39,922

Because it sounds like, I mean,

this, this tool has just been

00:42:39,922 --> 00:42:40,792

out for a little more than a

00:42:40,792 --> 00:42:45,082

year at this point in being

used by people, and obviously

00:42:45,087 --> 00:42:47,182

you're developing very quickly.

00:42:47,182 --> 00:42:51,922

It seems like I, I'm seeing you,

you're available on a new platform.

00:42:52,102 --> 00:42:56,812

You've got selective replacements, you've

got, you know, this new layering system.

00:42:56,812 --> 00:43:00,832

You've got all of this stuff that

the geometry, you know, adhering

00:43:00,837 --> 00:43:02,242

to the geometry versus not.

00:43:02,247 --> 00:43:05,422

These are all things that have kind

of shown up over the last year.

00:43:06,142 --> 00:43:08,542

What was a really valuable

piece of feedback that you got?

00:43:09,012 --> 00:43:12,792

Along that path that that really

kind of maybe even changed how

00:43:12,792 --> 00:43:15,462

you thought about the direction

that the development was gonna go.

00:43:15,824 --> 00:43:17,474

Benjamin Guler: I had like an

answer for the first part, and

00:43:17,474 --> 00:43:18,464

I'm not sure if I have a sentence.

00:43:18,614 --> 00:43:19,904

Let me, let me come up

with the second answer.

00:43:19,904 --> 00:43:23,774

But for the first one, I think the biggest

valuable feedback which we're working on,

00:43:24,134 --> 00:43:29,234

uh, even with our layer system is just,

uh, having control over the materials.

00:43:29,264 --> 00:43:32,264

Like having more control with

the materials and, and let's say

00:43:32,264 --> 00:43:35,084

if I want that to be wood, what

color would, should that be?

00:43:35,624 --> 00:43:40,094

Uh, specifically, so kind of getting

a bit more another level of finesse,

00:43:40,124 --> 00:43:44,204

like, I want this product specifically,

and so it comes in these three colors.

00:43:44,324 --> 00:43:46,634

Now there could be slight

variations because of lighting.

00:43:46,694 --> 00:43:50,774

We will always kinda change the diffuse

color of that, but that's something that's

00:43:50,834 --> 00:43:54,824

very evident that we, we see a lot of

times like, can we keep that material?

00:43:55,159 --> 00:43:56,959

Uh, um, you know, more finesse.

00:43:56,959 --> 00:44:01,399

And then I guess, uh, uh, another

comment on that one is, uh, being able

00:44:01,399 --> 00:44:05,569

to reproduce the same image more than

once, which is kind of making it be

00:44:05,569 --> 00:44:07,429

used more as a traditional render.

00:44:07,609 --> 00:44:11,869

So if I wanted to change the angle ever so

slightly, can I produce almost identical?

00:44:11,869 --> 00:44:14,869

And that's also something that we're,

we're working on to get, get that better.

00:44:15,289 --> 00:44:17,179

Um, something that

changed the development.

00:44:17,989 --> 00:44:19,579

Uh, I'm trying to think.

00:44:20,869 --> 00:44:24,979

I think, I dunno if I have a specific

one, but a lot of them along the way have

00:44:25,099 --> 00:44:26,479

altered the development of the process.

00:44:26,479 --> 00:44:27,859

Like from feedback that we've seen.

00:44:28,189 --> 00:44:30,469

Uh, we watch, you know,

social media a lot.

00:44:30,474 --> 00:44:32,534

We see people that are posting

on LinkedIn and, and, and,

00:44:32,534 --> 00:44:34,009

and Instagram and our forums.

00:44:34,009 --> 00:44:37,369

And so, uh, like we didn't think, I

think one time, like we didn't have

00:44:37,369 --> 00:44:40,819

like, portrait to be able to render

like decent portrait aspect ratios.

00:44:41,089 --> 00:44:42,139

So that was something, okay, yeah.

00:44:42,144 --> 00:44:43,459

That's something easy we could do.

00:44:43,459 --> 00:44:45,709

And um, so we added that.

00:44:46,009 --> 00:44:49,159

Uh, but yeah, that's, I think a lot

of smaller thing, you know, kind of

00:44:49,159 --> 00:44:51,259

decisions that were made that were, uh.

00:44:51,949 --> 00:44:54,694

That started to craft the application

more and more towards, you know,

00:44:54,834 --> 00:44:56,254

how, how people are using it.

00:44:58,054 --> 00:45:00,814

Evan Troxel: Can you talk about kind

of the, because this, this operates

00:45:00,814 --> 00:45:02,374

very differently than a traditional

00:45:02,374 --> 00:45:03,844

CPU or GPU based

00:45:03,844 --> 00:45:04,414

renderer where it's

00:45:04,414 --> 00:45:06,964

happening locally on the

hardware that I have, right.

00:45:06,964 --> 00:45:11,314

Versus this, where it's sending a

prompt out, there's an exchange of

00:45:11,554 --> 00:45:16,054

information happening, um, with the,

the model that you're using and the

00:45:16,054 --> 00:45:17,854

base image that I'm sending to it.

00:45:19,114 --> 00:45:23,344

Talk about like what that, what that's

been like for you because the, this

00:45:23,344 --> 00:45:27,004

is very different than the traditional

sense of like, what I need on my end

00:45:27,004 --> 00:45:31,084

to do a rendering versus a, a tool

like this and, and how you guys have

00:45:31,114 --> 00:45:35,614

kind of navigated that because this

is, this is a new paradigm when it

00:45:35,614 --> 00:45:38,284

comes to delivering images based on

00:45:38,404 --> 00:45:39,034

my model.

00:45:40,364 --> 00:45:40,744

Benjamin Guler: for sure.

00:45:40,774 --> 00:45:41,064

Yeah.

00:45:41,244 --> 00:45:41,464

So.

00:45:42,529 --> 00:45:45,619

The way it's set up is it's

cloud-based, the processing.

00:45:45,679 --> 00:45:49,009

So, you know, your model

stays on your machine.

00:45:49,009 --> 00:45:53,299

It doesn't use your CPU or GPU,

uh, on your local machine, but

00:45:53,299 --> 00:45:54,559

it gets in up to the cloud.

00:45:54,619 --> 00:45:57,259

We render it in our GPU

instances, and then we send it

00:45:57,259 --> 00:45:58,879

back to your, uh, local machine.

00:45:59,539 --> 00:46:02,359

Uh, so that's kind of the,

the main setup for this.

00:46:02,569 --> 00:46:05,449

Now we do have, uh, enterprise

clients that want to use this for

00:46:05,449 --> 00:46:08,419

more sensitive projects, and they

just don't want to have a connection

00:46:08,424 --> 00:46:09,439

to the internet while using it.

00:46:09,559 --> 00:46:12,229

And so we are working on

a, a local deployment.

00:46:12,229 --> 00:46:15,619

So then if, you know, if you have server

racks and you're your, uh, a company that

00:46:15,624 --> 00:46:20,389

has a few different, uh, GPUs that could

power it, then uh, that's also something

00:46:20,389 --> 00:46:22,759

that we've, we're working with a few

different, actually, companies that,

00:46:22,879 --> 00:46:24,709

uh, wanna have that local deployment.

00:46:25,129 --> 00:46:29,419

So what's nice about that is that,

you know, those companies that kind of

00:46:29,424 --> 00:46:34,339

need that sensitivity to, to explore

is behind their kind of, you know, um,

00:46:34,609 --> 00:46:37,099

intranet or their, their, uh, uh, system.

00:46:37,489 --> 00:46:41,149

And so that their networking doesn't

kind of, you know, leave that, um.

00:46:41,809 --> 00:46:45,859

So, yeah, that's how it's con And even

with that setup though, it's still not

00:46:45,859 --> 00:46:48,499

on your client, it's on a local server.

00:46:48,499 --> 00:46:52,009

So you have like a local server

system that, you know, load balances

00:46:52,009 --> 00:46:54,439

your requests and then sends back,

you know, whoever's using it.

00:46:54,529 --> 00:46:58,909

Um, 'cause it's really nice to not have

to have that, um, that on-prem setup.

00:46:59,149 --> 00:46:59,389

Right.

00:46:59,394 --> 00:47:00,259

On the client side.

00:47:00,499 --> 00:47:04,759

'cause then it's, it's like 30

gigabytes of just data just to, for,

00:47:04,759 --> 00:47:08,179

for one installation for example,

where you have that, uh, if you have

00:47:08,179 --> 00:47:10,939

it on, on your GP since then it's,

you know, it's more manageable.

00:47:10,939 --> 00:47:13,669

And then you have, you know, your

lighter clients be able to, you could

00:47:13,669 --> 00:47:16,789

even use, you know, um, a very thin

client to, to just do anything because

00:47:16,789 --> 00:47:19,699

it's just, you receive the images

and you interact with it that way.

00:47:20,499 --> 00:47:23,164

Evan Troxel: I I, I don't know if you

guys have done any kind of calcs based

00:47:23,194 --> 00:47:27,184

on that, but it seems to me like that

even is a, a, a valuable selling point

00:47:27,184 --> 00:47:29,944

to a firm who you, everybody's got a

00:47:29,944 --> 00:47:31,354

balance hardware spend,

00:47:31,624 --> 00:47:31,894

you know,

00:47:31,899 --> 00:47:34,504

capital expenses versus

this kind of thing.

00:47:34,504 --> 00:47:38,284

And to give people, um, the

ability to have, like you said,

00:47:38,284 --> 00:47:39,754

a really thin client, like to,

00:47:39,759 --> 00:47:41,674

you could have an iPad potentially, right?

00:47:41,674 --> 00:47:42,214

That, That,

00:47:42,214 --> 00:47:45,424

you are, you you're, you're

running sketch up on your iPad.

00:47:45,424 --> 00:47:48,304

You pull up a view and you say,

send this to, to Veras, and you get

00:47:48,304 --> 00:47:53,254

something back and, and you're not

spending $5,000 on the highest end.

00:47:53,854 --> 00:47:59,074

Laptop with GPU, you know, with an RTX

card, you, you've got an iPad, right?

00:47:59,104 --> 00:48:03,784

And, and you could effectively,

you know, in front of a client be

00:48:03,784 --> 00:48:07,954

coming up with this kind of prompt

based imagery, getting real time

00:48:07,954 --> 00:48:09,874

feedback from anywhere in the world.

00:48:10,264 --> 00:48:14,614

I mean, that's, that's gotta be a, an

interesting, that's gotta pique a leader's

00:48:14,619 --> 00:48:18,334

interest when it comes to like what

they need to outfit their staff with.

00:48:19,549 --> 00:48:19,849

Benjamin Guler: right.

00:48:20,659 --> 00:48:22,039

Yeah, it's, it's really, it's really nice.

00:48:22,039 --> 00:48:24,679

I mean, if you do have the resources,

you could do it and you can know,

00:48:24,684 --> 00:48:27,769

invest in the hardware and the

commercial grade hardware that we,

00:48:27,949 --> 00:48:29,629

we use like is pretty expensive.

00:48:29,899 --> 00:48:31,999

So you have to be very,

uh, like committed to that.

00:48:32,089 --> 00:48:33,169

Okay, yes, we're doing this.

00:48:33,529 --> 00:48:36,319

We're gonna buy a few of these and

it's gonna stay in our temperature

00:48:36,319 --> 00:48:39,589

controlled server room and, you know,

uh, everyone's gonna connect to it.

00:48:39,859 --> 00:48:42,499

Uh, versus, yeah, like you said,

like, um, you know, if you use the

00:48:42,499 --> 00:48:45,349

cloud instances, then it's, you

know, it's, it's very convenient

00:48:45,649 --> 00:48:46,279

to not have to,

00:48:46,579 --> 00:48:47,179

and, uh,

00:48:47,539 --> 00:48:50,239

you know, have that kind of

hardware, uh, you know, locally.

00:48:50,784 --> 00:48:51,647

Randall Stevens: do you have any stories?

00:48:51,647 --> 00:48:55,667

Who's, you don't have to say who it was,

but who's created the, what's the number

00:48:55,667 --> 00:48:59,627

of images that you know, somebody has

generated a single person with this?

00:49:00,582 --> 00:49:01,377

Benjamin Guler: Uh, I don't know.

00:49:01,377 --> 00:49:02,742

So we do collect analytics, but it's,

00:49:03,402 --> 00:49:09,642

but it's an anonymous, so we don't know

who, which is obviously, uh, but again,

00:49:09,642 --> 00:49:11,172

I haven't monitored that recently.

00:49:11,172 --> 00:49:14,952

I'm sure maybe Bill or, or someone

else will know more, uh, Ben Brun.

00:49:15,492 --> 00:49:19,152

Uh, but yeah, I think

it's, I, I'd have to check.

00:49:19,272 --> 00:49:20,442

I think it's in the thousands.

00:49:20,802 --> 00:49:24,072

Yeah, I think, I think per month

we get, I'm trying to think.

00:49:24,072 --> 00:49:24,132

I.

00:49:25,092 --> 00:49:27,222

Like half a million, maybe

renders or something like that.

00:49:28,032 --> 00:49:28,482

It's pretty,

00:49:28,962 --> 00:49:29,592

it's pretty fun.

00:49:29,592 --> 00:49:29,862

Yeah.

00:49:30,372 --> 00:49:30,612

Yeah.

00:49:30,612 --> 00:49:34,332

So we also have our free trial too,

so people kind of explore that.

00:49:34,392 --> 00:49:35,802

Uh, so they play with that.

00:49:35,802 --> 00:49:39,552

So a lot of users, um, you know,

um, have fun with it that way.

00:49:40,111 --> 00:49:40,286

Yeah,

00:49:40,541 --> 00:49:43,386

Evan Troxel: is there anything behind

the scenes that you could maybe talk

00:49:43,386 --> 00:49:48,366

about here when it came to how you

thought through the process of, of this?

00:49:48,396 --> 00:49:50,466

You mentioned you might have

some mural boards or something.

00:49:50,466 --> 00:49:53,496

I'm just wondering if, if there's anything

back there that you could show that kind

00:49:53,496 --> 00:49:56,046

of talks about how you handle internal

00:49:56,051 --> 00:49:56,371

development.

00:49:58,341 --> 00:49:58,611

Benjamin Guler: yeah.

00:49:58,611 --> 00:50:00,051

I was thinking about the beer boards.

00:50:00,111 --> 00:50:05,571

Uh, I think more in the context maybe

of, of local deployment and that set up.

00:50:05,571 --> 00:50:08,841

So I don't know if that is, you know,

that that interesting right now.

00:50:08,841 --> 00:50:11,661

But essentially, like, I think I already

talked about it, like it's essentially

00:50:11,666 --> 00:50:13,221

you have the client and then you have the,

00:50:13,761 --> 00:50:16,221

the instances, um.

00:50:17,436 --> 00:50:17,796

Yeah.

00:50:18,126 --> 00:50:20,076

Dev process wise, we're,

we're a small team.

00:50:20,196 --> 00:50:24,546

We're uh, nine, uh, we kind of,

our, our dev team is kind of

00:50:24,546 --> 00:50:27,576

structured like a Scrum team, so

we do agile, um, we have sprints.

00:50:27,846 --> 00:50:33,126

Um, we, uh, have internal releases

more often than, than, 'cause you

00:50:33,131 --> 00:50:37,806

know, some of these initiatives are

a bit, uh, more, uh, uh, ambitious.

00:50:37,806 --> 00:50:42,096

So we don't have a, we haven't had a

release, uh, a major release, actually,

00:50:42,096 --> 00:50:44,916

no, we did, we had the vector one very

recently, so I shouldn't say that.

00:50:45,126 --> 00:50:47,256

We had, that was, that was, that

was, we didn't have one in January.

00:50:47,256 --> 00:50:47,616

I think.

00:50:47,706 --> 00:50:50,136

Uh, we always have patches, so we

always kinda released a smaller patches.

00:50:50,136 --> 00:50:52,176

I think last year we

had a lot of iterations.

00:50:52,176 --> 00:50:56,166

Like almost every month we had

like a, a, a really big release.

00:50:56,316 --> 00:50:59,466

Uh, right now, I think one thing

that we've worked on from the dev

00:50:59,466 --> 00:51:03,396

side, which is a challenge, is, uh,

uh, just like working with certain

00:51:03,396 --> 00:51:07,326

applications that, um, are a bit

harder, like harder APIs, uh, for

00:51:07,331 --> 00:51:08,946

example, Vectorworks as a c plus plus.

00:51:09,186 --> 00:51:10,566

So that's like something that like.

00:51:10,851 --> 00:51:15,591

You know, is, is a harder language

to code in, and then documentation

00:51:15,591 --> 00:51:16,971

might not be the best documentation.

00:51:16,971 --> 00:51:19,101

So, uh, but it was, it was really great.

00:51:19,101 --> 00:51:22,671

Like it was a very good collaboration

with, uh, with, with their team.

00:51:23,061 --> 00:51:27,081

Um, we also have like SDK that,

sorry, I'm kind of blabbering on.

00:51:27,081 --> 00:51:29,421

I'm just trying to think about,

uh, kinda how we're set up

00:51:29,421 --> 00:51:30,501

and what we've done recently.

00:51:30,921 --> 00:51:36,111

Uh, where we, with our SDK, we actually

implement Veras as a separate application

00:51:36,111 --> 00:51:37,521

within your web app, web application.

00:51:37,551 --> 00:51:40,161

We're working with, with a few

different, um, parties that

00:51:40,161 --> 00:51:44,361

built their own, uh, formal like

application that they use internally.

00:51:44,481 --> 00:51:48,231

And so if you wanted to, you could

add Veras as an extension to that ac