From Sketch to Shelf: Transforming from End-to-End the Entire Visual Pipeline with AI

Vivatech

Transcript

00:00Hello and it's really nice to see so many nice faces on the stage and thank you very much for

00:08organization of course to invite us on VivaTech conference and I'm really happy to show what we

00:15are doing and probably to explain how we can change approach in a fashion content creation

00:23process and in a product creation process so we are Junira I'm CEO and founder of Junira and my

00:32background is a hyper realism so last 20 years I'm working in a different companies like in a CEO

00:40and founder to produce the hyper realistic content for a different industries like in an architecture

00:48film industry Hollywood etc and something around two years ago we decided that we can collect the

00:58best people from our industry and create new project that will bring hyper realism to the

01:05fashion industry too so and we put together experience from the fashion industry from a

01:12technology from a Hollywood production and from a game production put together and create Junira

01:19Junira team and our goal right now is replace physical photo shooting process video photo shooting process

01:28on the AI shooting process so last two years we are discovered technology and we are have a spoke

01:38with them probably more than 100 different brands so I am personally made the different calls with the

01:45Calvin Klein Hugo Boss Karl Lagerfeld and another one brands and we're talking with them and ask what is

01:54the best problem right now they have and what is the best issue that they will want to figure out

02:01in the fashion

02:01production and we understood that the fashion industry completely have no any product or any tools or any

02:10technology for a visualization so for a production the product fashion industry need to create a sample make a photo

02:20shooting process of the sample and after after only after this long process merchandising team marketing team can

02:30understand product works or not in another one industry everybody using visualization process to visualize product

02:38firstly and after make the decision and our solution it's a software it's a platform that put together technology of

02:51visualization to create a visual content for a product on the different stage so and you can see on the

02:59slide that we have a

03:00we have five main model that can cover the biggest and the most important part of product creation for a

03:09fashion industry so the first of all we are found a really great technology how we can work with AI

03:15fashion models on legitimated way i'll explain a little bit later how it works also we have a model that

03:22which is can create product detail page content lookbook creation campaign creation and also we have five capabilities

03:30So we have a better version of the tag, which can convert technical drawing or 3D to hyper-realistic images.

03:41So, and now you can see the video.

03:49That would be great to have a video.

03:51Oh yeah, we have.

03:53So, and on this video, you can see the first model of PDP creation tool,

03:59which is we already finalized and I believe that very, very soon, in two or three weeks,

04:05we will be ready to show this model to our first client.

04:10So, and you can see that the process is very interesting and it's very simple.

04:14So this looks like a shooting process in a real photo shooting studio, but with an AI.

04:21So the consumer or fashion brand can just upload the photo of the clothes

04:26and system will generate a lot of different images with this clothes

04:31and after our director team or marketing team can just choose the best one images

04:38for a future production, for again, for a merchandising, for a team, for a marketing team,

04:44for a social media, et cetera.

04:47And the work that you can make with the physical photo shooting process

04:52can be compressed in 10 or even 100 times

04:57and we can make it much more easier, faster, cheaper with the same quality

05:04because you can see that the quality of the final image is amazing

05:09and the final product can be finalized in a five key resolution.

05:14So, and this is just the first model from five models

05:18and very, very soon, I believe that in two months,

05:21we will be ready to show to the world the model that can make the campaign production too.

05:26So, this is the first model and as I tell,

05:30it will be finalized in a few, just in a few weeks for a world.

05:35So, on the next slide, you can see the quality

05:38and quality, this is what's the most important thing for us

05:43because when we are spoken with a brand,

05:46we realized that there's not enough generate 99% of the quality

05:53if you are talking about the details, et cetera

05:55because we got a lot of feedback from the fashion brands

05:58and they tell us there's not enough quality,

06:02the detail is not enough,

06:03you should improve the hyper-realism, et cetera

06:06and we spent something around 12 months

06:09to generate images, to train the model,

06:12to create our own data set

06:13to reach this kind of quality

06:15because now every stitch, every button,

06:18every detail on the right place

06:20and also we are working not only with the clothes,

06:24we can work with the accessories

06:26like glasses, bag, shoes, et cetera.

06:30So, it's a real step how we can optimize production process with an AI

06:37and it's not only about the content production for a marketing

06:42so you can imagine then product team can also make the visualization

06:46before going with the sample production.

06:50So, on this slide, you can see how we can reduce the cost of production process

06:57and time of production process.

06:59For example, average price for a PDP production can be between $50 and $300 for each image.

07:08In our case, our client will pay just $1 to $5 for one image.

07:13Campaign creation, so it's super expensive process

07:17and fashion brand can spend a lot of money for campaigns

07:20so it can be $100,000 of the euros.

07:23In our case, one shot can cost from $10 to $15 only

07:27and take much less resource and of course time.

07:31So, we can compress time for a content production

07:35also in 10 in even more times.

07:38So, that means that the process of delivering new collection on the market

07:45will be even in two times faster and easier than before.

07:49And I think it's a very important slide

07:52because we understood that our platform

07:57it's not just a platform for a content creation.

08:01So, this is the platform which can connect all departments together

08:08because fashion, it's about the visual, first of all.

08:12And when product teams start working with some of the collection

08:16they can place images on a general space platform

08:20and give access to merchandising team, for example.

08:23And so, that's mean that we break the bridge

08:28between the product team and merchandising team

08:31merchandising team and marketing team

08:33all of them can work in one space

08:36and be in one visual content.

08:39So, that's mean that product team can discuss

08:41with the merchandising team new collection in a few days

08:44not in a few months

08:45very quickly and very efficient.

08:49So, next one slide and next one point

08:52is very, very important for me

08:53because when we started working with AI faces

08:57we realized very quickly

08:59that we can't generate faces from an AI only

09:04because it's unlegitimated.

09:07And we thought, okay, what we can do with that

09:10because fashion industry should use AI fashion model in production

09:14but each face should be legitimate.

09:17And we found the solution.

09:19As I tell, I have a huge background

09:20in game production and movie production

09:24and we found the way

09:26in which one we solve the problem.

09:29So, our faces is not AI faces.

09:32Our faces is a 3D face

09:35which our director is doing in 3D software

09:38like in a character for a movie, commercial, cinema

09:42and after we are using AI

09:45just to add the hyper-realistic approach for a face.

09:48So, what does it mean?

09:49Does it mean that each face

09:51has a 100

09:54legitimated IP rights for each face

09:57which is super important

09:58and each brand can avoid the risks

10:00that we are stalling some of the faces.

10:02It's a critical point.

10:04And on the next slide

10:05you can see the real use cases

10:07how fashion brands

10:10use our technology.

10:12So, for example,

10:13this is a great example

10:14how we can convert

10:16tech pack

10:17yeah

10:18we are using tech pack

10:20we are using

10:21mood board from our client

10:23and after a system

10:25can generate the final image

10:26it's a PDP

10:28and so we have a really great use case

10:31with a brand Karen

10:32which is use only AI images

10:35AI content on the website

10:37for a launch

10:38the collection

10:40the product

10:41and you can see this on the web right now

10:43even

10:44also

10:45it's a really great example

10:46with our friends

10:47and our partners

10:48daily paper

10:49so because Genera can also work

10:51with the physical product

10:53so we can scan the product

10:54and create the final

10:56PDP or lookbook images

10:58and you can see that

10:59the quality is amazing

11:00and also

11:02I like this example

11:03because

11:04this example can show

11:06that VR works

11:06not only with a

11:08drawings

11:09or with a tech pack

11:10we can also work

11:12with a 3D model

11:13so each client

11:14can upload

11:153D model

11:16of future clothes

11:17system will convert

11:193D to AI sample

11:21and after

11:22we can make

11:22the campaign creation

11:23for example

11:24or something else

11:25and on the next slide

11:27I think that

11:27the quality

11:28this is the most important

11:29part of Genera

11:30and on the next slide

11:32you can see

11:33the quality

11:34of our production

11:35and

11:36yeah

11:37this is the PDP

11:38this is the lookbook

11:39and this is the editorial

11:41which is also made

11:42with our technology

11:45and

11:47it's a little bit

11:48complicated slide

11:49but I'll explain

11:50so

11:51on the top of this slide

11:53you can see

11:54regular calendar

11:56which is

11:57each fashion brand

11:58super brand

11:59like a tier 1 brand

12:00using for

12:03creation collection

12:05and the period

12:07it's a period

12:09for a collection

12:10creation

12:10it's 12 months

12:11that's a lot of

12:12so if we will put

12:14Genera

12:14on the

12:154 main

12:17stages

12:17of the

12:18collection

12:19creation

12:19we can compress

12:21this time

12:22from 12 months

12:23to 6 months

12:25and now

12:26we have

12:26some of the partners

12:27that are already

12:28using this technology

12:29and cut

12:31every month

12:32cut more

12:32and more

12:33and more

12:33time

12:34for

12:34content

12:35and

12:35collection

12:36creation

12:37so

12:38where we are

12:38right now

12:40stage 1

12:41on the stage 1

12:42so it was

12:431 and a half

12:44years ago

12:45when we are

12:46attract 70

12:47brands

12:48make the conversation

12:49and all of them

12:51begin

12:52our

12:52design partner

12:54based on

12:55the feedback

12:56we are

12:56go to stage 2

12:58and we are

12:59proof

12:59with all of the

13:00brands

13:00quality

13:01so now

13:02we are

13:03on the stage 3

13:04and some of

13:06this brand

13:07is already

13:07using our

13:08technology

13:09and very very

13:10soon

13:11during the

13:112025

13:12and 2026

13:13we should

13:14make the

13:15adoption

13:16of our

13:16technology

13:17for 20

13:18slash 30

13:19percent

13:20of the real

13:20production

13:21for each

13:21brand

13:22so this is

13:23how we are

13:24going

13:24and how we

13:25are developed

13:26right now

13:26so I believe

13:28that very very

13:29soon

13:30probably in

13:31one or

13:31maximum

13:32two years

13:32each

13:33whole industry

13:34will convert

13:36not whole

13:37production

13:38from physical

13:38to AI

13:39but big

13:39part

13:40of commercial

13:41part

13:41of course

13:42I understood

13:43that the

13:43important images

13:44like a

13:46campaign

13:46should be

13:47produced

13:48with the

13:49photos

13:49and with

13:50the real

13:50models

13:51but at the

13:52same time

13:52fashion brands

13:53has a lot

13:54of commercial

13:55content

13:56which we

13:57can optimize

13:57with an

13:58AI

13:58so this

13:58is our

13:59mission

13:59I believe

14:00that you

14:01like this

14:01presentation

14:02and thank

14:03you very

14:03much for

14:03your attention

14:04if you have

14:05some questions

14:05please ask

14:10I'm coming

14:11I'm coming

14:12thank you very

14:12much

14:13I'm waiting

14:13on you

14:13so before

14:14we wrap

14:15it up

14:15because you

14:16did an

14:16amazing

14:17presentation

14:18thanks a lot

14:19of

14:19and we're

14:20super excited

14:21to use

14:22Genera

14:22I think

14:23you already

14:23shared with

14:24them

14:24when it's

14:24coming out

14:25did you

14:26of course

14:26yes

14:27so how

14:28they can

14:28follow you

14:29because I

14:29think

14:30Genera has

14:30a lot

14:31to share

14:32and if

14:33people want

14:33to really

14:33keep up

14:34with the

14:34updates

14:35what is

14:35coming

14:35where

14:36they

14:36can

14:36find

14:36you

14:37it's

14:38a really

14:38good

14:38question

14:39so you

14:39can

14:39find

14:40us

14:40with

14:40the link

14:41generaspace.ai

14:43and

14:44I believe

14:45that in

14:46two three

14:46weeks

14:46we will

14:47launch

14:47the first

14:47version

14:48of the

14:48software

14:48okay

14:49yeah

14:49and

14:50we will

14:50make

14:51the

14:52news

14:52in

14:53our

14:53social

14:53media

14:54and

14:55after

14:55some

14:55of

14:56the

14:56iteration

14:56we will

14:57contact

14:57the

14:58media

14:58like

14:59Evoke

15:00and

15:00another

15:00one

15:01of our

15:01partners

15:01to

15:02explain

15:02the

15:02world

15:03that

15:03we

15:03are

15:03ready

15:03to

15:03launch

15:04the

15:04big

15:04version

15:04of

15:05Genera

15:05amazing

15:06thank

15:07you

15:07so

15:07much

15:07thank

15:07you

15:08very

15:08much

15:08congrats

15:08thank

15:09you

15:09all

15:10thanks

Catégorie

Transcription

Commentaires

Recommandations