How the FLORA Museum Challenge turned Met artifacts into a runway show

Lightly edited from the recording for readability. Wording is the speakers’ own.

Hi, hello everyone. Welcome to our latest FLORA live stream. I'm your host, Sasha Zabegalin, a forward deployed creative here at FLORA. I primarily work with a lot of brands that we serve, particularly helping them onboard with our tool, teach them the tool, consult on a lot of their AI projects and initiatives.

But in my personal time, I specialize in fashion and footwear with AI. Any questions about that? But moving on, today's live stream is all about our latest challenge with Contra, the FLORA Museum Challenge that is running until September 7th. Today, I'll be live building a submission to hopefully spark some inspiration for you as you submit for this challenge and hopefully become a winner. I'll be one of the judges so excited to see what you all cook up and I hope today live building with you all can help you get started and create something beautiful.

But before we dive into building, let's take a little overview of the challenge. So yeah, today is a whole live stream dedicated to our FLORA Museum Challenge with Contra. It's an open design session with me where I'll be ideating an idea one at a time until something blossoms. My canvas is-- OK, cool. Let me make sure my canvas is displayed on the live stream.

So yeah, I have a deck on here. But yeah, the challenge is to pick a piece from the Met, from their open public archive, find some things that inspire you, give you a bunch of ideas, and then either go ahead and reimagine it through things like time, material, scale, placement, and its purpose, or if you'd build a world around that piece from the Met. There's over half a million works inside of the Met's public archive, so there's a ton of inspiration to pull from for creating something beautiful. And of course, the submission for the challenge is to enter by September 7 at 11.59 PM Pacific time. And so when you submit, again, there is the FLORA Contra Challenge link right here. I'll paste it in the chat so you can check it out yourself. And feel free to check out that page.

That's where you'll submit for your challenge. But the rules are to pick-- make sure you pick an open access piece from the Met's archive. Follow the brief of reimagining the piece itself or building a world around it, all inside of FLORA. Make sure you publish your project to the community page, and then share the link to that as well.

And then in your submission, make sure you also link the original work and a short note on your inspiration itself. Finally, post on X or Instagram, and be sure to tag FLORA and use the hashtag #FloraMuseum. There's over 10K in prizes, which is super cool. The first place winner gets $5,000 and a paid expenses trip to the Met here in New York. And you'll probably come visit us at the office too, so be pretty fun to meet you in person.

This gets $1,500. Third place, $500. A FLORA favorite, picked by one of us judges, can earn $500. And then the fourth through eighth place winners win a year of the FLORA Pro Plan.

So super exciting. We're judging based on originality, craft, storytelling, and also its museum ability. You can sign up for this challenge for free. You can submit up to three entries for this challenge. And if you don't already have a FLORA account, you can create an account through that control link that's in the Control Challenge page itself, and earn 50 FLORA credits on us to participate.

So with that, find the piece that stops you in the tracks and create something beautiful. Yeah, let's get started on the challenge. So before the stream today, I actually went on the Met's page itself. If you just look up the Met Archive collection, you'll be able to go on their site and find this.

But something important to know, make sure that you click has open access here in the search button. This is just to make sure you can freely use these images as the Met has granted free distribution. For the sets of images. So, super important. You use these as your reference points.

But what I did before today's stream, I just went on here and just searched blue because everything I create is in this blue futuristic aesthetic that I so love. And I just searched through a bunch of stuff and just pulled out what I found to be kind of inspiring for me or visually interesting. And so they're all here. And what I'm going to do now is just download all of these. Save image. I'm going to make sure to download all of these. Save image.

A ton of really cool stuff like I really have. I'm really winging this. I have no idea what I exactly want to create. I was thinking something with fashion inspired by some of these pieces or bringing something to life. I don't know. I'm still there's a lot of really cool objects that I found inside of the Met gallery.

So I don't know, just going to pull from there and kind of just see what happens with AI and just kind of go from there. A lot of I really love Van Gogh growing up. My mom is a huge Van Gogh fan. And everywhere in her house we have Van Gogh paintings or memorabilia and books. And I remember always as a kid browsing through them. My mom was an oil painter.

So she made a lot of oil painting pieces that were very reminiscent of Van Gogh. So I think there might be something there to create. There's fashion pieces as well inside of the Met collection. So a lot of stuff I think done there.

So I'm just going to save a bunch of these quickly and just go through them and then throw them all into the canvas and see kind of pull apart what's speaking to me and think through some interesting ideas as well. Potentially even creating some videos. Because C dense 2.5 that dropped like a week or two ago now. And it can take up to like 30 images, which is really interesting. A lot of interesting potential there. We also dropped fable 5.1 yesterday inside of FLORA.

So I'm thinking of using that as like a text node to help analyze the pieces and like create some concepts based on that. Just go through these. And again these are all open access pieces from the Met. So safe to use and copyright free and able to redistribute it. Someone asked a question regarding Fauna. Is there a way to turn off auto mode? I haven't quite sure.

I haven't used Fauna in a minute. I believe you can. It used to be. But if you use Fauna, definitely recommend checking out our MCP. We've done a large upgrade with our MCP and you're able to open in agent now and have chat with Claude open and FLORA canvas open at the same time. Yeah, you can do a lot with them.

OK, I think I'm on the last one. Let's download this. Download image. Someone asks, would you explore them individually? I don't know. That's why I'm going to attempt. I'm thinking of either combining some of them, exploring them individually.

Just depends on how I feel through the process. Okay, I think I have, yep, I think I have everything. So I'm just going to dump all of these images onto my canvas. Let me pull up my computer.

Okay. Just going to drag all of them onto my canvas. For the guidelines we're making, making use of all. I'm just going to explore a bunch of concepts. I'm not submitting anything because I am a judge.

So it's more so, just giving you ideas on how to do things. But, yeah, the submission guideline is you can only submit up to three pieces. So, choose your three strongest. Things you like to submit. That would be my recommendation. Really try to build something that's, really interesting and tells a great story and the visuals feel, on the level of the, really embody that. And we'll explore some, what are the best, image models and stuff to use for, a lot of, a lot of this process.

That's well. Okay. Come on files. Download please. It's taking a while. Well, okay. I'm gonna keep this going.

Maybe try one more time. My internet's being kind of spotty. Okay. Well, I'll let it upload for a second.

Okay. Damn. It's stuck internet. One second. I'm going to switch my wifi. Hopefully it doesn't.

Okay. Testing. Hopefully we're back. Let me know in chat if you still hear me.

Okay. Okay. Cool. Yeah. My streaming software said that it's, it like ended.

So I was Oh no. Okay. Well, there is all the pieces. Let's, I'm just gonna group this together just to make it distinctively clear. I'll label this blue and call it pieces.

Okay. Well, let's think through what could we do? There's a lot we could do. Well, I'm going to try.

Okay. Why is it? Okay. There you are. I really liked these Egyptian things. I feel like there's something I can do with this.

Okay. Yeah. Let's take out all that Egyptian stuff. Let's, I don't know.

Let's build some fashion. Pieces around these. So, okay. I have these three pieces. I'm going to connect this to a Texan and have Claude help me kind of ideate a list of, prompts based on this.

Let's throw this onto fable 5.1. Cause it's the current smart, smartest language model in the market. Just curious to, pick its creative brain and so let's say something like let's say something like here's some inspiration pieces I've grabbed from the Met Archives from the Met Archives they have me thinking of creating some fashion pieces inspired by this collection of objects and their creative direction based on this and your experience in a set of roleplay prompts so it writes a little better and your experience as a senior fashion and footwear designer craft a list of image prompts for a full body high fashion for various full body high fashion let's say Egyptian and mummy models wearing full body pieces including interesting footwear including interesting footwear all set on a plain white background front facing shot front facing shot front facing shot okay output only the list of prompts okay so I'm going to have fable analyze these three visuals from the Met and I'm giving it prompt like here's some inspiration pieces I've grabbed from the Met Archives they have me thinking of creating some fashion pieces inspired by this collection of objects and their creative direction based on this and your experience as a senior footwear fashion and footwear designer create a list of image prompts for various full body high fashion Egyptian and mummy models wearing full body pieces including interesting footwear all set on a plain white background front facing shot output only the list of prompts okay so I will have fable cook up on that let's let's see what it thinks it's gonna create take it a little bit while that's doing its thing let's prepare with a batch node so I can just quickly run through these list of prompts and visualize them I'm gonna use the create to image model because I feel like it's really fun at specifically text to image because right now I'm just gonna try text to image and just see how I feel so okay we got our batch of prompts I just click input split by line so I can go through each individual prompt and then connect this batch into an image node let's switch the model to create a create too large I'll keep its creativity to medium for now let's do a four by five just so it's like an Instagram post ready I'll run this I will duplicate Let's go to Krea2 medium. Let's also set it to 4x5. I'm just curious how Krea2 is going to think of this. Well, that is and I'll also try.

Let's try actually connecting these as a reference and. Actually. Krea2. References large.

Let's connect this. And let's run this. I'm going to use the Krea2 references model and plug in these visuals because it's going to pull stylistically from that batch. Of stuff. I'm just curious what's going to think at the same time.

Let's also run nano banana 2. And nano banana pro. Let's run that. Let's run C Dream 5. I'm just going to throw everything at this.

Let's just let's see whoever does the best will be my winner and I will take. Those assets and run was it. GPT. Image 2. And if you are ever stuck on like top image or video models to use, I've personally went into our floor admin page and set the featured models to be the ones I recommend for most of your tasks.

And then a pro man to check image to see Dream 5 Pro. These are the top. Image models. I would say personally.

OK. So let's. Go through. I kind of like that's pretty cool. I don't like this one that looks weird. Not bad. Oh.

That's cool. I don't think nano or C Dream or any of those are going to do too well. Personally. Yeah. Like they're fine.

But I think the problem with these models is like they're more for utility stuff. I call them utility models versus the best sort of concepting text to image model. There is it's it's the Krea2 models. Mid-Journey in FLORA. And they just have like a Genesee quad to like look and feel like these visuals are so much more visually appealing in my eye than like these. And create two references didn't do too bad but you can see it like pulls it very literally from the design style. Oh Lord.

That looks so tacky. I don't like any of those. But OK. Let's keep browsing.

Let's see what we got. Oh that's fun. High fashion money to go. OK. Come on.

OK. All my zooming is going crazy right now. Sorry. It's kind of fun.

Let's just browse through. Interesting. Let's see if any of these. I hate all of these. Oh Lord no.

OK. Well we can give those a chop. We'll stick with the Krea models specifically to just concept out some pieces. I'll do a couple more. I want to create a few of these.

And then when thinking of this maybe I'll go into the archives and try to find some sort of setting and then have these models do like a runway walk in that world. It's pretty interesting. Okay. I think what I could do is maybe let's switch the model to, let's use Opus 5. I'm just gonna rerun the same prompt. Just have it think through that problem 'cause every text model you use, every single different language model you use, kind of thinks differently.

So it will read my prompt and think through that topic in a different approach and manner. So that's how I can kind of keep the same theme but have it kind of brainstorm some other variations and takes. It's like having a different designer take a look at the brief and come up with their own sort of idea. So let's do this. I just duplicated these nodes and what I'm gonna do is change the creativity level of the Creator2 model. Meaning it will either, if you set it to high, it will loosely follow your prompt. It will like in essence kind of follow it but not fully.

So it can invent more interesting details versus if you set it to low, it will more strictly follow the prompt. So I do this just to kind of see how different creativity settings have an effect on the image output that we're going to get. So I'm just gonna rerun these. This will be my last run and then we'll keep moving.

But yeah, so far I like these pretty cool pieces. Let's see, what else we got? I actually went to the Met for the first time like two weeks ago. And I remember seeing that Egyptian exhibit the first time in real life and it was just mesmerizing. You feel an energy there for sure. There is a essence there. You can like feel the ancient history and like the people is just really cool.

This looks awesome. I love that. Okay, let's browse through. Interesting. Actually, I kind of like this one.

Okay, let's keep looking. Let's see. Ooh, that's fun. That's cool.

Okay. Let's see if there's anything else. Okay. I think we got some good stuff. I think I'll roll with this one over this one.

Okay, so we have a bunch of pieces. Let me just group these together. So they're nice and tidy. So we have our first step. We've created some, in this case, for my submission, I'm thinking of analyzing these objects and creating a collection, a fashion collection inspired by those objects.

And then now we have our collection ourselves. And so what I'm thinking now is maybe let's find some like backgrounds to place these models into. I want to create like a catwalk inspired by either a real Egyptian work or some painting from the Egyptian past. Okay, there's a lot of objects, a lot of objects.

Let's see. Let's see if we can find something I feel like I saw Okay. Let's see, let's see. I feel like I just saw some sort of Egyptian set up. Where was it? I forgot what the piece was called.

Okay, I think it was on highlights. It was in highlights or something. If you're working on the challenge, feel free to leave in the chat. What are you thinking of? What are you dreaming of? What is inspiring you? There is a ton of really interesting things.

I feel like an interesting submission I would love for someone to see is like making a video game mock-up of one of the various different settings that kind of exist. There's a lot of objects inside of the archive. As you can see, here's a sword, here's a vase, here's armor. I feel like you can put all those pieces together into a video game kind of idea of inspired by these real pieces of work.

Okay, I genuinely cannot find this. Okay. Oh, okay. Oh, there it is. Found it.

Let's see if I can get one more, please. I might have to just roll without one. Oh, wait, there's only three pages. So let's let's see if there's anything.

Okay. Well, did not find that it was I want to use the right terminology for this. So okay. Temple of Dendor.

Okay. Let's go back here. And let's just upload my reference of this background. Okay, so we have, I'm just gonna, I'm just gonna note this location down for myself.

Okay, just so when I write the prompt, I have a better sense of context of what to say. Okay. Let me group this pieces. Okay, so next step is let's build like a catwalk pulling from this image.

So what I'm going to do is I rely a lot of people ask me about prompting and best prompting practices. It's utilizing the text note. This is the same as if you had Claude or chapter BT open on your computer by yourself. And you have like FLORA open or another tool open, you ask it to like write a prompt for you or like a really good example is when you use Claude or chapter BT or Gemini to write an email for you do the same sort of thing inside of FLORA. It just is this node based canvas.

So just visually looks a little different, but same sort of idea. So in this case, I'm just gonna plug in my reference image of this temple of Dendor into a text node. I'm gonna use Claude for this example. I'm gonna use fable 5.1. You can see it's attached.

So it's able to see it has eyes on that visual. So I'm gonna write a prompt like or I'm gonna have it role play, so I'm gonna say as a fashion, runway, stage, designer, expert. You're given this reference image of the, let me copy and paste this, up to the temple in the Met Gallery inside in New York. You were asked to come up with runway concept renderings set at this installation. You're asked to come up with runway concept renderings set at this installation. How to put only the prompts. As to come, as image prompts.

Okay, so I'm having it analyze this image and say as a runway stage design expert, you're given this reference image of the temple to endure in the Met Gallery in New York. You're asked to come up with runway concept renderings set at this installation as image prompts. How to put only the prompts. I'm going to just throw that idea at Fable and have it come up with a list of runway concept renderings,. And just let's take a look at what it thinks of.

But the idea is I'm going to create a runway inspired by this piece. Then I'm going to have all of these models catwalk that runway. And yeah, we'll see how it goes. Okay.

Then I'll batch all these together. Then I will connect this to an image node. Let's make it 16 by nine. So we get a really kind of a widescreen view of this runway and take a look at what it does. I'm going to duplicate this four times so I can run the same prompts with the same reference image across the top four consistency models. In this case, the reason why I'm using the consistency models over Krea is because I'm trying to actually pull from this reference and have it actually use this scene that I'm inputting into it as like kind of like the world to be in versus the clothing stuff. It was more so just analyzing it and coming up with in with words of how to describe the new visuals and kind of come up with new fresh ideas versus this one.

Way that like actually looks like it was like it's here in New York in this installation in real life. While that is running, I actually just have an idea of creating a scarab looking actually I have an idea to apply this texture. Let's remove that. Apply this texture onto the floor logo so I can use it as an end scene.

This campaign video is fashion runway. I mean, so I'm going to do that is I think I have a prompt saved that does something. Okay, I think this is yeah. Okay, so I have a prompt that I'm going to use just saying analyze these imagery and detail extract and describe the mood and tone the color palette the cultural references. The top typography and graphic language, the photographic and visual style, the product preservation, the key motifs. I'm just going to have an analysis step that analyzes both of these images.

And then from that analysis, feed that analysis into another text note that's going to apply that style onto a graphic. In this case, I don't know, I'll just keep throwing things at fable and just keep testing it out because I'm just curious of how well it does with creative tasks, especially since it just dropped yesterday. Okay. While that's cooking, let's see.

Okay, I actually don't want people, okay. So this is actually a great topic. So a lot of times you'll get a result from your image models and you don't like something about it. I usually just go back to the prompt and add in those stipulations. Like for example here, I mean, I don't really like these runaways actually, but there's people in it. I don't want people inside of this yet. We're just trying to create a blank slate.

So I'm just going to go here and type that in. Okay. There should be no people featured in these renderings. Someone asked, how did you create this set of models? I created a set of models in the beginning of the stream, but found these images in the met archive, connected them into Claude. And I had Claude come up with a list of prompts. I just said, here's some inspiration for this.

And so Claude came up with a list of prompts. And then I just connected the list of prompts to, a text to image model, I created two, which is like a journey in FLORA. And it has this very kind of really interesting results it likes to do, which are really cool that I aesthetically enjoy. So pick those out, pick my favorite ones. And now we're moving on to creating the stage on which this world is going to be set in. Something I actually want to do now is actually realize these renderings so far. Just what I'm picking up on is it's fine, but I don't, I hate these.

I don't feel like it ties in cohesively with this story world and building. So let me add that into add that as context into Claude. So then it also picks up on the aesthetic sense. As well.

Okay. There should be no people featured in these renderings. Let me actually make this a little bigger. Let me make this a little bigger so we can read it a little better. I don't see people featured in these renderings. The other attached images are the creative direction, mood board pieces to follow so that this stage which you'll set in the same world as them. I'm going to have it now come up with a list of new prompts.

All that is generating this analysis step finished. Now let's use Opus 5 in this case. I'm going to say based on this analysis, craft a list of prompts that apply. Texture to this logo in the same style. In the prompts mentioned to use a logo as reference, only the texture, the logo, should change. I'll have to put only the prompts. I'll have that do its thing.

While that's doing its thing, we have our list of prompts that Claude has created for this runway concept. For these runway visuals, I will run them. And see what happens. Hopefully they come out better than what they were. I did not like any of these looked very sloppy and not chic enough. Just kind of boring in my view.

Okay, so let's do that's going to take a little bit. We got our list of prompts. Let's do that. Let's batch them out.

Let's connect our image here. Let's go there. Let me duplicate four of these really quickly. Let's do that.

Let's switch this to Nano 2. Nano Pro, GPT Image 2, and C-dream is there. Let's run all of these so I can apply this sort of texture look to our logo. I'm thinking of using this as an end frame for the video visual I'm trying to build up towards. That's awesome. Take a look at it. I'll be one of the judges.

Okay, let's see. The redeem code should be in the description of the Contra page itself, and it will explain more there as well. Okay, this one's not bad. Okay, I'm going to try. I'm going to try one thing. I feel like maybe by connecting this image as a reference to these models is making it too much too that image, and I actually don't want it to stick that hard. I'm just using it as inspiration.

Let me take that off and just see what happens. All that's happening. Let's also use some models that don't take in specifically. Reference can't take in images. They're only text to image models, because some of them are really good at creating awesome visuals. Let me throw out ideogram four. Let me throw out re-craft.

I know re-craft's pretty good. Re-craft. Oh, no, let's do 16 by nine. Let me throw in re-craft. Let me also throw in, what's another good one? I think Reeve could be a pretty good one, right? Reeve, yeah, let's use Reeve.

Let's also throw in, let's throw in Kreia too large and see what happens. That's actually kind of cool. If you've been here before, there is this pool in front of this temple, actually. So that's kind of cool. I actually kind of like this one, not too shabby.

Okay, I'll do one more run of these and then decide and try to finalize. Okay. Actually. Okay. The essence of this cat walk should be chic. Minimalism, art as fashion, couture angle, something Anna Matur would want to be at. All of the shots should be front.

Facing angle, the cat walk itself. Given the layout with a list of I'll let it do its thing. Let's have that go. Okay. Well, that's running.

Let's see. Cool. Wow. Okay. I'm going to pick out which ones I feel like hit the mark.

Let's do that one. Let's do this one. Okay. Pull those out.

Let's visualize these. The stage, this is my final run of trying to find a great stage. I don't feel like any of these stage renderings so far have hit my personal mark for what I like to see. All of these are fine, making me go whoa quite yet but the logo texturing pretty cool I think I'm gonna use this one that looks really cool take a peek oh this is not bad yeah let's let's see anything I didn't see anything at seed dream did so let's take our GBT image did so let's take a look okay I think my I think my top one is actually gonna be this one I'm just gonna use this one okay so now that I have all of my pieces I going to just lay them out here let me group these so it's easily labeled and I will share this project with you all after as well just to hopefully give you some inspiration stage renderings okay let me duplicate this put that there let's copy this put that there so we have all of our bits and elements of things now let's bring it all to life so what I'm going to do now is hopefully can I title this okay so I'm going to have now Claude come up with a prompt that blends all these things together and we're gonna throw it into Seedance 2.5 and create the cool video out of it so connected all my visuals into Claude here and now let's let's let's come up with something okay senior action creative director concepting runway show or your collection inspired by archival chip eat chip Shin I spoke that Egyptian which is firing Egypt. A spell check and some Egypt. Shen by archival Egyptian art pieces. You're a senior fashion creative director, concepting the runway show for your collection inspired by archival Egyptian art pieces.

You have eight pieces to show on this catwalk you've designed. You have eight pieces to show on this catwalk you've designed. Filming this. Make sure I'm using table 5.1.

This runway show. You're filming this runway show. You want the models to walk down this runway. In their own dynamic flare that makes sense with the garment they are wearing. They're wearing. Filming style should. Feel.

Almost. And. Held. Like someone in the audience is courting and broad. Casting this runway. That's set at the. They pull up.

Oh. What was it called? Temple of. Let me pull up the title so I just make sure in the prompt itself I'm using the right terminology.

OK. That's set at the temple door at the Met. NYC. You then want this. Video to end with. The logo. In.

Textured stone. Output only. Now. Create. The prompt. She's.

This. Vision. Outfit. Only. The prompt.

OK. So. I. Have all my kind of Lego pieces on my building blocks now I'm going to use the power of Seedance 2.5 which can take in a. Bunch of reference imagery and kind of blend it all together to create a really awesome video especially with Seedance 2.5. You can generate videos at up to 30 seconds long so I can create a really long video with this which is really cool.

But what I'm going to do is have Claude help me write the prompt that achieves the vision that I have in my head. I just connected all my inputs all the assets I already created. And using Claude I'm saying I'm giving a prompt like you're a senior fashion creative director concepting the runway show for your collection inspired by archival Egyptian art pieces. You have eight pieces of show on this catwalk you've designed. You're filming this runway show. You want the models to walk down this runway each in their own dynamic flair that makes sense with the garment they're wearing. The filming style should feel almost handheld.

Let me fix that. Handheld. Audience is recording and broadcasting this runway that's set at the Temple of Dindoor at the Met in NYC. You want this video to end with the floor, logo, that section, and show. Overall, this video should feel chic and worthy of the Met. And, yeah. Thank you, Jim, whoa.

It's been, I don't, I wish, in like your Nintendo or PlayStation or your Steam, how when you play a game, it displays, it tracks on the hours you've played on that game. Wish we had that with FLORA, because I feel like I don't even know how many hours I have. Inside of this, my brain kind of thinks in node graphs now, which is interesting. You want this video to end. It should be 30 seconds long.

OK, I'm going to have Claude cook up on that. OK. Oh, cool. OK.

So now I have a prompt. I'm just going to drag all my reference images first into the video node. Then let's connect my runway and connect my textured logo. And then let's connect the prompt. We have Seedance 2.5 already preselected because I'm just using this all the time.

And then let's set the duration to 30 seconds and let's set the resolution at 1080p. It's going to cost 36 dollars to run one of these videos. And we're going to do four because we are on the FLORA land. And, yeah, let's let's take a whirl at this. Oh, Lord. Well, that's running. Let me take a take a look at the chat.

Someone asked a question, I have a current student received a code with credits be added at some point. I'm not in charge of that and I'm not quite entirely sure. But again, I think the guidelines are in the description. The guidelines should be inside of the Contra page. You can leave any questions or comments there and our support team will help you with that. Andy said, I love to see different creative processes and that each person has when working. These kind of videos are my favorite way to see what I can readjust and incorporate into my workflows.

That's great. That's what we hope to do with these live streams kind of show, the creative process of what it's like to work in for a lot of times, it's a lot of times when we do stuff, we already have the workflow done. And so we just present the final result and like the final kind of thinking and approach that we did fix camera. And so it's fun to show what it's like behind the scenes and how long it kind of takes to imagine and create things. Yeah. Oh, yeah, we need the achievements and make flora feel like a video game. Actually, kind of fun.

Maybe we do that. I'm just going to have I'm going to duplicate Seedance and run it with the other prompt I gave. I'm actually going to prompt this one more time. And this prompt out like a film director would describe. I'll let that run. Prompting is always a challenge, but the cool part is that you can use AI to help you with prompts. And that's usually what I do most of the time.

I just have a vague general idea of what I'm going to do. And then I have AI kind of take a pass in a stab at that vision. And then if I don't like how AI portrays it, then I go back and add some more context and stipulations to my original direction. And then try that again.

And then see if I like it. I like it. Then I roll with it. Sometimes AI can give me something better than what I was thinking about.

So that's always fun to see. Okay. I'm going to just duplicate Seedance 2.5 one more time and have it run this attempt as well. Cause each prompt, all three of these video prompts are different. And so I'm just curious on how it's going to portray what I'm trying to achieve. Hope it's good. Please.

Usually Seedance 2.5 does create some of the craziest videos I've seen. Oh, wait. I do actually, this is actually a good test. Oh. I know we have a new model I just dropped. It's the minimax.

Okay. I'm actually going to try because I know Seedance is going to take a while and probably get the best result out of it. But there is this new model that's been like breaking the internet that dropped on Friday of last week. It's called minimax H3 max. Some of these are very crazy names, but it generates like at lightning speeds. It's the first, like almost real time video model where it's just like super quick at creating visuals. It's super quick at creating visual, creating videos.

And so because I know this Seedance is going to take a hot minute to finish, I actually am curious of what the visual could look like from minimax. I don't know. I haven't tried this model quite yet. Okay. How many pieces? One, two, three, four, five, six, seven.

Okay. Seven pieces. Let's connect all of these. And supposedly it's as good as mid-journey or Seedance. I don't know.

Let's just take a stab at it and see what happens. Let's generating. I'll take a look. At any of these are probably going to take a really long time because they are 30 seconds and tiny resolution.

So it's a big reason why I don't make too many videos. I make a lot. I feel like I make a lot of imagery stuff. So yeah, because the videos just take a really long time. I just feel like I'm sitting and waiting a ton versus like images. I feel like I get very quickly. And I can blaze through a bunch of ideas there.

Okay, let's give Minimax a whirl. Let's set it to... Let's see. All that happens, actually. I want to show something cool. We've added our 3D node officially.

So you can do something like taking... You can input an OBJ and some other 3D files onto the canvas, but you can also access 3D models to turn an image into a 3D object. I believe from what I've been seeing, the best ones have been this HI3D one and also MeSHI. I heard was really good.

So let's turn this. Honestly, let's turn all of these. I think this is my favorite piece. So maybe let's see if I can turn this model into a 3D object. I'm still figuring out the use cases for a 3D node, but I think so far what I think it's fun at is turning any image into a 3D model. I mean, they're not the greatest 3D model you'll ever see in the world, but they're pretty decent now. I've used AI3D model stuff way back in like 2023, 2024, and at the time they were terrible.

So they've progressed pretty significantly, but understandably, I think the MeSH and the subsurface textures in this aren't like AAA level, but it's good at creating a 3D model really quickly of something. If you need to use it to prototype something in like a 3D blender or cinema or something, or Unreal or Unity. Yeah, it's a hive mind for here. Yeah, I use a lot of credits here at FLORA. I don't even know what my usage is. Oh, I can't see it. That's strange.

But I think from August, I think I reached up to like $7,000 of spend. But, we got to work with clients and they got to build a ton of stuff for them. Okay. Well, I don't know where that failed. Let me try something. There's always something with Seedance.

Okay, well, let's see what Minimax has on this. I mean, not bad, but I Seedance is gonna do way better at this. I already know. Like what the hell? I need to generate quickly. There's that. Seedance, please generate.

I just want to see what you think. Okay, I feel like my beginning content blocked. I'm gonna say do not mention the met in the prompt. I think Seedance is probably blocking because there's a lot of content restrictions. Because, white dance got sued.

So they have a lot of restrictions they've put on Seedance. I'm thinking I think the prompt the video prompt itself. Mention the met and I think it's probably flagging that and causing a failure. So sense of graceful style that makes sense. Carmen they're wearing. Give it Let's hopefully. At the same time, let's I don't know if I actually like that handheld.

We run this. See what happens. Because I don't know if I actually like the I don't know in my head. The handheld look seems interesting. I think Seedance will probably do a better job at that than this mini max model per se.

But still, I think potentially in execution. It's not going to turn out how I'm thinking because oh, I kind of want like a make it feel like it's like someone recorded like live streaming on an Instagram or something like they're at the show and like filming it. But I feel like the videos that it's going to produce aren't going to feel that good comparative to like saying not mentioning that and having it just kind of guess and think through what it could be. We'll see.

Okay, I'm just gonna run all of these at the same time. Pray that one of them spawns. Please. Oh, I just broke down for Okay. While that's generating, let's take a look at our 3d objects actually.

Let's go. Let's go full screen. See. Well, not bad. We got our little fashion piece as a three object.

Let's see how that was messy seven. This is h1 3d. I feel like the lighting is a bit better on this one, but the accuracy of character, I think Meshi did much better at that's cool. And then here is our floor logo. Oh, that's awesome. Well, we have officially got a 3D printer in the office now, so maybe I'll 3D print this object, actually. Yeah, I can download it as an STL file if I wanted to, so I could take it into my 3D printer.

That's cool. I don't know. Let's turn this into a 3D render, too. Let's just have some fun while this goes. Anything at all. No. Maybe let's turn this mummy into a 3D object while we wait for the videos to generate.

Just duplicate that and use meshi7. The really cool, I think one of the coolest features of our 3D node specifically is if you have a 3D object, either generated or on the canvas itself, if you like import it from cinema or Blender, wherever you make your 3D objects, you'll see at the top right here, you can create a sheet of images of all the various different angles of the object pretty quickly. So for example, if you're like a footwear brand or a brand that has packaging or goods and items, and you have 3D objects of whatever your product is, you can import it into the canvas and extract all angles of it really quickly to turn it into an element, per se, or use it for context as for something else. So you just quickly get all angles of said object.

Okay. Please. Anyone generate a bag? What file can you download in after 3G model? Right now it appears to be JBL, OBJ, STL, OBJ, STL, PLY, and USDZ. And we're still rolling out updates. And there's a lot coming actually, which is why we haven't made a big public announcement of 3D node on canvas yet because still perfecting it, but it is out and available.

So feel free to explore. And if you have any feedback at all about it, you can leave feedback actually by going to this help and resources icon right here with the question mark and go to suggest a feature. And you can suggest any additional updates you want to see in for the 3D node or anything else inside of FLORA as well. Okay. Well, all that's going to take forever to generate.

Let's take, I'm going to see if, oh, wait. Seedance 2.0 can't take in that many references. I pray. Please. Seedance don't fail me now. Yeah. Well.

Take a look. Let's move this back here. Yeah. While these are generating, feel free to leave any questions in the chat about anything really and I can have some live Q&A time for you all while these take, forever to generate. Yeah. Also curious to hear of anyone's ideas for what they're building right now for this challenge and what they're currently thinking. I'd love to, think about it and talk about it.

So feel free to share any of your ideas in the chat of what you're cooking for our challenge. And yeah, I'm excited to look at all the submissions and see who the winner is later next week. I'm going to try toning down the resolution to 720p and hoping that's going to help with the speed of the Seedance generating. Well, maybe when that's generating, let's take a stab at, I don't know, something else. I'm trying to think. I mean, I love that Van Gogh painting.

So let's take a stab at, I don't know, applying some sort of motion. That could be kind of cool. An idea is just let's create a video where the painterly textures of this Van Gogh's painting like feels alive. It's moving. Create a prompt that animates. Painting as a living scene while maintaining its originally painterly look and style. There should be what is it?

Texture. Surfacing. Oh my God, they're all failing. One second, I'm going to reach out to our Eng team and see if they can actually help me diagnose the problem. One second. And stupid five. Going to our engine team.

So maybe they can help me find the problem. Ensure. Okay. Someone asked what are common things to avoid that you see with the comm submission. I would say originality. I think I want to see something interesting kind of same sort of kind of process of how I was approaching this challenge instead of like I see a lot of submissions where people are just it's the same exact art piece just like I don't know in addition to it or like I don't know if they like imagine the Mona Lisa but she's like in the galaxy and I'm just like no it's you to approach the challenge as like you are an artist in her using these visuals as like a reference point to build a world on top of rather than like use that exact image and then like change something about it's not super creative but like oh I was looking at this paint piece this like paint piece of thing whose flowers that made me think of creating a bunch of like ceramic objects based on that or imagining various plants in the same style or sun dresses inspired by Van Gogh's flowers or something just like taking that inspiration image and like building something more extravagant on it. Yeah.

And I also would say for submission definitely recommend recording a video of you speaking about it. I think it's always interesting to hear like your ideation and thinking process then just a visual that way it gives more context and breath to the work itself. Did we get something? Wow. That's so cool. One of them to spawn and it was the one I sent to the 720 not bad. I'm curious to see what other ones there are but I'll still I'm just going to upscale this video just so I can get a 4K version and at 60 FPS so we can get a really nice thing there.

I'll let that cook. Yeah. Let me have this play. Okay, a couple of questions in the chat. Someone asks, "Sasha, while we wait, could you quickly walk us through how your process started and thinking behind it just to wrap up everything, please?" Yes, I can. Just one second. Someone asked questions, "Does it have to be a video or an image counts as well?

That being said, will the video be rated more?" It's images or videos, whatever you feel. I just wanted to create a video today because videos are just engaging for people to watch, and Seedance has gotten really powerful at creating awesome videos. We made some cool visuals alone, but let's create a world around it. Let's make it more living. It's up to you, whatever you prefer, if you want to make a video, sure. If you want to make imagery, sure. You can have a collection as a submission, and that counts as one.

For example, one of my submissions can just be these garments inspired by those Egyptian objects, and that could be by one submission as a collection piece. Definitely encourage a collection of stuff because I feel like it brings more life to your images. Let me see if I can make Seedance spawn out one more anyway. I'll do the same.

But recapping over my process of today's live building for this challenge, I first went to the Met Archive. Make sure you toggle on this. Access image, that way you're using copyright-free imagery from the Met's collection of stuff. I just searched up for my case. Everything I do is in a color blue, so it helps really steer my creative direction. I just typed in blue and just scrolled for like 30 minutes, just finding a bunch of interesting visuals from the Met's archives that I found intriguing. I just opened up a canvas, dropped everything in there, and then I just randomly thought of using these Egyptian ones because they really stood out the most for me.

And then using those three images, I connected it again to a text node, and I used Claude in this case, and I'm using Claude to help me write prompts,. I have an idea. I just told Claude, "Here's some inspiration pieces I've grabbed from the Met Archives." They have me thinking of creating some fashion pieces inspired by his collection of objects and their creative direction. Based on this, in your experience as a senior fashion footwear designer, create a list of image prompts for various full-body, high-fashion Egyptian and mummy models wearing full-body pieces, including interesting footwear, all set on a plain white background, front-facing shot, output, only the list of prompts.

So Claude gives me a list of prompts to try to visualize a collection inspired by these pieces, because Claude has eyes on the pieces itself, so it's going to think of that world. And then I batch them all together so I can quickly run through all my prompts. Before something like batch, I would have to do split lists in the nodes and isolate each prompt myself. It would take a really long time to then go image and generate each one of these. It's going to just take really long, but with batch, I can just batch all the prompts together and then run them, run through them by just connecting out an image node and selecting my model and going. I use the Create A2 image model. It's a really great text-to-image model.

It's mid-journey in FLORA. It has that same sort of aesthetic sense, where I can create really cool things like this. I find it way better than Nano and others, especially for fashion looks, because I feel like it reminds me of mid-journey, where I can actually create interesting visual fashion pieces rather than generic-looking things. So just ran my prompts through Krea and visualized some pieces. I picked out which ones were my favorite.

Then from there, I thought to myself, let's create the runway that I want these models to walk through. So I found an image from the Met Archive of the actual temple of Dendur and also connected my reference images of the objects that I'm using as the creative direction for this runway. And then I asked Claude again, I said, I connected all my visual references, so it has eyes on it. So when it's writing, I was asking it to write. It sees what I see.

So then I'll write in that sense of direction. So I prompted it as a fashion runway stage designer expert. You're given this reference image of the temple of Dendur in the Met Gallery in New York. You're asked to come up with a list of runway concept rendering set at this installation as image prompts. There should be no people featured in these renderings. The other attached images are the creative direction, follow this so the stage feels set in the same world as them. The essence of this catwalk should be chic minimalism, art as high as a fashion couture angle, something amateur, when tour would want to be at all of the shots should be front facing angle of the catwalk itself, output only the prompts.

And then again, connected my batch of prompts to an image node and I used the four consistency models of Nanobana2, Nanobana Pro, GPT Image 2, C Dream 5 Pro. And I tried some other text image models as well. Picked out which ones were, which stage renderings were my favorite. At the same time, I also then created a textured FLORA logo, just having Claude analyze these imagery and pull apart its characteristics right here.

And then connected that analysis into another node where I fed in our FLORA logo. And I said based on this analysis, craft a list of prompts, apply a texture to this logo in the same style in the prompts mentioned to use the logo as reference. Only the texture logo should change output only the prompts. And then we got a list of 10 prompts that apply this sort of textured logo on it, just so I can kind of concept out some various different looks. And we got our FLORA logo. And that sort of texture that we pulled out of these. From there, I have all my building blocks necessary so I can throw them into a model like Sea Dance and have it create some cool campaign videos.

Let's see if anything will spawn. Yeah, then the final step was using all the building blocks that I did of the earlier stages. And now connecting all those building blocks into another text node where I'm having it help me write a video prompt for a catwalk animation featuring all of these items, all these models and the stage and the end scene together. And so far, we've only gotten one thing to spawn. And now I am upscaling it so I can get a nice because this generated at 720p. It up to 4k so I get a really nice visual. If nothing spawns out of the other results with Sea Dance, I'll just call it a day with this one because at least we got one result out of Sea Dance.

And we can just roll with that. Yeah. Let me see what-- because this seemed to have worked. Cool. I hope I'm saying a name, but I hope that covered things for you, Maria.

And then up here at the top, I'm just exploring our image to 3D node that we have on the canvas. And not too bad, actually. This one turned out pretty sweet. While we wait for-- I'll give it a couple more minutes while we wait. Any questions so far I can help with, whether in flora or models and anything? Let me group up these assets, and I will share the workflow link so you can try it out yourself.

But I will know if you submit something from my workflow, because I was the one that made it. So do not try to submit something I made, because I will dock you for points for that. Creating a runway. Creating a logo as texture. Collection. Concepting. Art.

Pieces. Let me make this all nice and pretty. OK. I will share the only link. Yes, we're working on actually adding and animating the 3D models that you make here. That's one of actually the next updates that will be coming to the 3D node itself, which is really going to be really exciting. Come on.

Let's see if I have a-- no, I think a Magnific upscaler. I love this video upscaler, but it does take four. My live building is always a challenge. So if you're still tuning in, I appreciate you for sticking around, because live building can be kind of boring at times. We're waiting a lot. That's a lot of AI work is waiting and praying that the result you're going to get lands.

But it's addictive. It's almost like a slot machine of creativity. Oh, we did find a bug. And I'm just working on that. Hopefully.

So then probably I'll come back later today and like try to generate some more of these, but just trying to upscale this one at least so we can get one cool video out. I'll let those generate. OK. But I'll let the full workflow be in view so we can see whatever spawns out whenever it's finished.

OK. Yeah. I see you, Sophia. You can clone my project. It's in the right hand corner at the bottom bar. You'll see clone project and clone it into your workspace. Otherwise, you would only be able to view the project itself if you don't clone it.

So be sure to clone the project so you can work on it independently. Yeah. We have a ton of exciting updates that are dropping soon for that. Can't wait for you all to see as well. I think. Yeah. Please finish.

Please. I can't walk entry for the challenge. Oh, awesome. Yeah, I'm excited to see what you guys bring. I. Oh, that could be another interesting angle thinking of footwear inspired by some of these paintings. I think there's a lot of cool things that you could do with that, especially with some of the pottery stuff that I've seen in the archive itself.

I don't know. Imagine some actually let's actually do that's actually kind of cool because this is going to take a minute. I'm going to I'm going to do this under these video generations so I can immediately tell when something's done. I'm going to say something like. You're a senior footwear designer given this object as inspiration or. New. Sneaker collection.

A new way, not sneaker. Let's say, yeah, no, actually, let's do stinger for a new sneaker collection. And for a new sneaker collection. Create a list of image prompts, concepting. Design. It's content and design output only.

Let's go to a fable. Let's do that. Sure thing, if you, if a question, if you could use Fashion Studio help. If you start off with, oh, yes, you have to submit a workflow.

So you can start off in Fashion Studio, then bring into the canvas and start building on the canvas. Cause there's not a way to share a project from Fashion Studio as like a public link or anything. So that's just something to be mindful of. Okay.

So we got one more video. I wish the odd, you could hear the audio, but. Wow. Just crazy how, oh my God. It's just crazy how, to Seedance 2.5 is just wild with what you can do.

Okay. Here's the upscale. The old version of this video. Let's have it play and then we can end off today's stream with that. Cause we are an hour over actually. That was only supposed to go for one hour, but we ended up going for two. Let me refresh this page.

Come on. Well. Okay. Well, we have our upscaled video now, so I can download this video and it'll be 4k. And you can, submit something like this.

But something to note when I leave you off with, there's, we have a timeline editor inside of the canvas, so you can input your, CPA. Dance video, or any videos that you have. And then natively edit it in our timeline editor. It's like cap cut but inside of the canvas, for you to use as well. Let me. Yeah, it's like cap cut.

So you can I don't know, let's say you don't like this part right here. You just want it to immediately go up. I can just click cut there, do that, move my timeline there. Zoom into my timeline.

Let's say we want to add a fade effect so we can like fade into the scene. Things like that. So you can, go the extra mile with your video, edit it, make it pinpoint precise to like exactly the pace and beats and all the things that you would like. But yeah, hope today. Give you a great starting point and inspiration for how to approach your, FLORA museum challenge. Had a blast concepting with you all. Super fun to do and really excited to see what you all submit for this challenge.

I'll be one of the judges. So excited to see what comes through and yeah, have a wonderful rest of your day and, we'll hear back more from us next week and have fun creating all the things with our museum challenge till then take care and we'll see you next time. You you