Lightly edited from the recording for readability. Wording is the speakers’ own.
Hi, guys. Welcome. Good morning, good afternoon, good evening, depending on where you're calling in from today. Welcome to this week's FLORA webinar featuring Ariel Ibañez, Head of Creative AI at Superside. We're very excited for this one. Ariel has been involved with FLORA for a long time, so we're really happy that he's here today to share some of his work in FLORA.
Thank you, Olivia. It's a pleasure to be here in the garden again and I love this webinar, so I hope I make justice to every other creator. They are total legends. Today, this one is called AI Bloom and hopefully we are going to see and understand how we growth ideas with generative AI and FLORA, and we have been doing this in Superside for more than three years now.
We have 350 creatives that works fully with AI and creating AI systems on a daily basis, and we're going to try to give you some glimpse of what we do and how we do it, and we always want or like to start with why this and why now. I think it's a really, really nice way to understand why this is important on the industry right now and, to be completely honest- and this is something that I really hate- before, great ideas used to lose at least a week to do switching and you were going from this tool to another tool and then handing over things to other creatives- never was an easy task to do and down. We have one canvas and they don't. The ideas don't have to lose anything. They can be as good as they are in your head at first and you can scale and grow those ideas and we're going to see today how easy it is to do that with FLORA.
We have, and hopefully we are going to do this super fast on the presentation side and we're going to jump into FLORA to get this done live. And I'm going to show examples of things that we did try to discuss, also credits with you and times, so the amount of effort behind every project and the old way, the old way that we were saying it was one idea, at least six steps and we were having the concept of being a moodboard app. We were doing the scripts on a document, the storyboard on a totally different app and final images and videos and we were trying to merge all together and every hand off for us cost time because part of the idea and it started to dilute the original concept. So the idea treats a little bit more with each hand off, and it's not the way that it needs to be and, like we were saying, this is how we were used to work because of the limitations we had. But now, since things have changed in a better way and we have generative AI to help us, we can certainly have a better way to do it.
So we're going to use FLORA. Obviously, you mostly of this tool is one amazing canvas that has every model available, so you will be able to connect text, to connect image, to connect video models- everything in one workspace. You're going to be able even to edit those videos, and we're going to also have an editing part on the demo. You will basically drag a block and you can think about this block or this node as LEGO blocks connect to another, watch the concept, move forward from the brief and from words to pixel and to motion, and you are not leaving this app.
So this is the beauty of this canvas and basically, you have more than 50 models and we're going to leverage agents too. That AI agents, or how I like to call it, agentic behavior, is something that is highly trending right now and we are going to use in this case Claude, but you can use ChatGPT or any agent. You have to leverage that and to control the tool for you, and we are not going to switch from any tabs to do this.
We're going to use five phases on one canvas and this is something that we do for every project. We always start with a strong concept, a strong idea, and the ideation phase. That, to us, is highly important: how we deconstruct that brief, how we understand the context, how we understand the clients and the brief. And then we script and we are going to understand what we are going to do, in this case, a short video. So the script for the video, the times, each scene, how it needs to be portrayed.
Then we will craft a quick storyboard, frame by frame, on one style, keeping the consistency, always taking into account the restrictions that we have, all the requirements from the clients. After that we're going to go to the bigger phase, which is production, and I think also a beautiful thing about generative AI is that before we were jumping into phase to phase. So if you log in the ideation, then you will go to the script and then you do the storyboard, but you will never go back to the ideation phase and every phase was totally isolated from each other. Now, with generative AI, basically everything happens on the same file or can happen on the same file.
Nothing will get lost and we can jump between phases. We can say, for example, okay, if we are on the storyboard phase and we need to do some tweaks on the ideation or the script, we can certainly do that quite easy and fast and we don't need to be waiting one week for someone else to change the script for us. This production phase: we can certainly have one round, two rounds of production until we are happy with the output, and after that we will have the delivery phase when we will log everything that is on those assets and get the final things done for our clients.
On the ideation and concept phase, we here are going to drop a brief, a reference image. If we have a rough line of copy into the canvas, something to start with, of course, as much context as you have will be better, and then the ideations will propose move we haven't considered. It will understand the context and the client it will help us draft those ideas and those concepts in a better way. But we need to have our human insight, we need to give the agent something to start with. So we will start to connect the things, to get dozens of directions if we want, and this we can be here on this phase as long as we need to get these concepts and these ideas into a better way and make things that matters, that have impact, right.
So if the agent is going to suggest variation- but we can certainly move as we go Once we go to the scripting, we started to understand if there is copy here, voiceover, something like that, of course that should be logged in here on the scripting. So any voiceover shot descriptions, scene descriptions, anything that we need for that final video. In this case we are going to do 15, 20 seconds of video, so mostly three, four scenes, no more than that. But everything will be here and we can have again as many scripts as we want and as many back and forth with the agent as we want until we are happy with the result. And after that, moving to one of my favorite parts of the process, which is story mode, because here we are starting to see how this project is going to be realized and we are going to start seeing some visuals.
And this, to me, is when you need to hold every part of the consistency if you are working with characters, if you are working with some branding the product, for example. Today we are going to showcase some product shots and it's highly important for us to understand that there are still limitations on the technology. So most likely you are going to check consistency here, check that the product has the right scale, that the details of the product are showcased right. So everything here on the storyboard is one of the most important parts for you to go to production with a consistent set so you can reorder here, you can branch anything and you can certainly get variations of your images until you are happy with the result. And after that we get into production, we can turn those approved frames into motion with image to video blocks, and here we are going to use Seedance as our preferred video.
But there is a video model. But there is a lot that you can use. You can use MiniMax, Hailuo or you can use a lot of video models to test what you are doing and to leverage. Also, cost versus output. Obviously, compare everything in the place.
That's one thing that I love. Here you can see every result in the canvas, understand, attack them, put colors on them to see: okay, this needs to be changed, this needs to be tweaked still, or this is approved and you are going to iterate until you are happy with the result and then deliver to our clients. So you will refine in place. Right now there is no need for you to go to any other app, even for editing those videos.
You can export those finished assets. You can share the canvas if you want. You can make some order too and then approve anything, make comments, even in your canvas.
So the idea today is to do some live demo. So we are going to here live on live. We're going to build those five phases from one brief and I have a cool product to work with and I'm going to show you really quickly. I'm going to switch to our FLORA tab here and, as you can see, we have here our product. This product is a speaker with a visualizer, so it has four screens and it's able to visualize images, not in 3D, it's a faux 3D, but it looks really cool and it's a nice product to have.
It's called Kit and we have the brief here. I'm going to quickly read it so we can understand what we're going to do, and then we're going to connect the agents and start working in the canvas. So Kit is a smart speaker and light visualizer. It's a tall, slender tower of smoke translucent acrylic on a short metal base. Inside there is a grid of LED pixels that reacts to sound, and this is the beauty of this project.
Since those pixels react to the sound, we can create a lot of animation and our ideas can be translated into the same product inside those screens. So off it's a quiet black monolith and on it comes alive with pixel animations. It can reproduce equalizers, waveforms, ripples and even a pixel cat, which is part of the brand and part of the logo. And the base carries the cat mark and the Kit wordmark, both backlit, so you can see here the product off and the product on, with the cat animation similar as the logo shape. We have these two images as inputs.
So Kit off and Kit on and we have every image we make. It will be referenced from both and this is important. So this type of brief: when you work with AI, you will need to be super accurate when you talk to the agent to say, hey, we need those two images as inputs every time that we created a new asset, a new key art for the model to understand and to stay true to the real product. The objective of the campaign is to launch Kit as design object- it's not a gadget- and it shows the movement, the moment a silent option becomes light and sound. We have the audience here, we have the proposition, the copy is sound, you can see.
So we're going to base on that to create our campaign and the tone will be premium, minimal, quite confident, with a little mischief, and our direction will be minimal too, like Apple, like diagraphite, similar studio. And then we have some other considerations here. Due to time I'm not going to dig deep much into those.
But finally the deliverables are three key arts from one campaign, same look, of course, and we have the hero studio on a modern day. So setup: we have the same setup as one, but someone is turning the product on. And then we have same but different camera angles as key art. And finally, the 15-second product, film.
So this is what we are going to work and how we are going to build. This is same as we were talking just now: five phases: ideation, script, storyboarding, produce or production, and then delivery. So these five phases are here.
So you can see I leave a place for the agent to work. So what we're going to do now. We have these as inputs and also some examples of previous campaigns that we did. So you can see here the product in a key art already done, and then we have here the logo as an asset and the space for the agent to work. So this is quite easy to set up.
You will go to open in agent. You will go to Claude, open, Claude, and here we have the setup message that FLORA has a standard. So we're going to write some other things here. So we're going to obviously use the FLORA MCP and the model will understand the file and we'll read the brief. So we are going to just start with read the brief on the FLORA canvas and start working on our phases.
I'm going to paste here, just in case, the phase ideation, like we were saying, script, storyboarding, production and delivery, and then we are going to do some things as notes for the model to understand. So it basically has screens but doesn't project anything. So it doesn't project anything outside its screens and that's as one. We can also put some valid notes here. We're going to say, for example, don't focus too much on the cat animation, that's already covered.
We have done this before, so we don't need that and we can say something on our idea for this video and this 20,. Let's do a 15-second video, fast-paced with a bomb animation. Let's say something like that to start with. So the agent gets ongoing, we're going to press enter here, check that everything is connected, that the agent can access the file and understand what we are requesting, and once this is starting to work- and we are happy that the agent is working, we are going to showcase some examples too. So let me here understand what the agent is doing, so open a specific project and it's trying to read our file.
In the meantime, we can go back to what we have here and I will showcase. Let me check the agent just now. It says that I already read the brief, the back shot and the logo. Next I will write ideas and script on the canvas, then get the storyboard frames ready, to quote one thing. Is that an important thing?
I'm going to stop this right now and I'm going to do another note to the agent just in case, because we are on a live demo and this agentic behavior on Claude can be long and can work a lot. So we are going to just say: try to do this fast these first phases, fast, do one and two and come back to me. Don't use more than 10, let's say seven minutes, six minutes. Let's push the agent far, as far as we can.
So this is because agentic behavior can create basically a workflow with, let's say, seven agents, that the agents are trying to validate what the other agents are doing, and this can take a long time, not in comparison with traditional workflows, of course, but can take more than 20 minutes that we have for this event. So the idea here is to showcase how to get this done as fast as possible. And while the agent is working, I want to show you already these projects are already done and I will try to explain the amount of efforts, the hours and the credits that we use for this. So, as you can see, here we have the same thing with a brief, same behavior, same project, same inputs and everything. And on the ideation phase, the agent- in this case it created eight explorations plus nine screen animations.
So it started to give you ideas on how this product can create those animations, what can be inside them, and then different type of shots which you can totally curate or the agent will create it for you. The agent will start to read everything that he created and understand which ones work better. You can even have one of those agents to test this on a performance perspective. So from all of these animations you can see some of them are already going into this red filter, which is a limitation of the technology right now. With some models they get this red filter on top.
Once you iterate all a few times on the same image, that's not a problem. We can color grade the images here inside the canvas, which is also some really nice things to see. You can see here one color grade image already, so that's something you can state to the agent as well when you are creating this. So after the ideation phase, the agent creates the script, so we have the script for the final video and then the story board with the shot that it was created. This whole patch, like you are seeing here, this whole five phases with the ads as well, took less than 40 minutes, 35 minutes and only 78 dollars on credits, which is pretty much fast and cheap for this level of quality and this accurate campaign that we have.
So in this type of cases, the best case scenario is that you make a stop after the storyboard and you curate the content yourself, you can start to understand what works and what doesn't work. That well, besides the color grading that, like I say, that's something we can fix in any part of the or any phase of the project. Understand if the close-ups are good or if you want something different, if you would rather want, let's say, an on-product close-up, if the animations looks accurate, if the scale of the product that we were saying it's accurate, because in this case it looks a little big in comparison to the real size of the product, and so on, you can start understanding or, for example, in this case, showcasing some video games, which is not part of what this product does. It can showcase the animation, but it will be pointless to just showcase this and then not being able to play it as a video game. So we are going to discard these images.
You can tag things. I usually put red colors or green colors if the image is rejected or approved. After the storyboard we get into production phase and we started to get more images because we ask, we say to the agent after the storyboard phase: hey, listen, we need more images, I need a desk, I need a computer, like the brief save. Sometimes these details can not be here from my ideation, but again, you are always the human insight on each phase. It's really nice to have a quality control checkpoint on any phase and then move on.
After that we have the final shot. So we can see here how the agent is part of the video and obviously here sound is something that we're going to take into highly consideration during the process. So you can start creating these short animations and these short videos with sound, or you can remove completely the sound and create the sound after. Again, another beautiful thing about the AI: you can have a lot of things going on and at the same time.
After this, we started to ask the agent for editing those scenes and to create the ads. You can see here our six final ads and we have some feedback as well for the agent. But in this case we just move on and color grade the two best options, put that into production, understand the performance and then we were going back to a second batch.
This is one of the final videos that the agent created. I have volume, so let me showcase this really nice video correlated with music. It's perfectly synced, the animations looks good, the product looks on scale. We have nothing but good comments about the video. Like we say, there are some ads here that got rejected mostly because of scale.
Here the product looks like figured and it should, but that is again something that we can iterate after. But in this case we didn't have to do that. We just approved this one and move on. So you can see, this was first batch. Like we were saying: 35 minutes, really quick, only 78 dollars on credits and we could pitch something to a client super fast.
Then we move to a second exploration: same brief, same input, a lot more of exploration. Here we started to ask twice as much of key art and we started to push the models and the agents to the bare maximum and start to understand how far we can go with those animations. And then we asked for a 30 second video instead. So these three scripts are for 30 seconds and they are going to start to get on the storyboarding and on the production phase. We're going to start to get twice as much footage as previously on the other batch.
So this a second batch. For you guys it will be easier. You already understood in this case working with the first quick exploration on 35 minutes. We already understood a lot of things about how to ask what we need and what to expect from these models and these agents, and after that I'm not going to dig a lot on this exploration, but after that we have a new set of ads, new set of images and a new video. Here I'm going to in this case, like we're saying, 30 seconds long, twice as much footage.
We still have some comments to tackle. That is not a final video, like on the round one, and these are not the final ads. We still need to work a lot on this and this exploration. Since we have- you can see here- a lot of rounds on the production phase. I think that one really great insight is, like we were saying, you can have a lot of time exploring ideation phase.
You can have a lot of images on the storyboard. The phases can differ in quantity and in time. In this case, we wanted to spend more time on the production phase- creating footage, editing and then to deliver a longer video with better footage, and we still have footage that we didn't use, so we can keep editing that video.
Everything that you just saw. I didn't touch one single note. The agent handled that more operative side, and then let's see how this is going, let's go to Claude and let's see what Claude said to us and we are going to go quickly and approve this. So it says: ideation and script are both on the canvas. Has no cards on one inside each frame.
Nothing has been generated. Oh, no. Ideation has three concepts. I recommend a, so to for us to select concept. I'm going to say concept a: short fuse, a single pixel spark light.
Concept b: come down. Concept c: hot potato. The light bomb hops from phase to phase as the camera orbits. It's fun, but it needs one look long orbit take, which has more rigs to generate. So I'm going to go with the suggestions.
I'm going to say a here and then I'm going to read the second question that Claude has for us. So script is short fuse, 50 seconds, which gives eight bars and a cue, and I cut every two beats in the montage. Okay, perfect. I'm just going to say a and Claude is going to start generating for us. Hopefully this is going to be fast for Claude and we can see how the model and the agent is starting to use the canvas for us and we can approve and curate the content.
In the meantime, let me open FLORA again and we have here the notes on the model and the agent for you. So it will document everything that you do if you put feedback and will create the note and, like we were saying, this is a nice way or a really good best practice to have things are approved so you can tap the images, a screen or even if you want to have more color on the canvas and not much of a color person, but you can tag the fields green too if you want the model to understand, say, hey, everything that is stuck green, it's approved. You can already see here that the model is connecting, without me doing much, connecting to directly to the storyboard. It went from ideation script, storyboard in this case didn't create images on the storyboard and it's going to already create the prompts using both images as input.
As we were saying, we need the on and off as inputs to understand how this product behaves and for the screen to be accurate as well. We cannot rely just in one input in this case and go back to the agent and I will just say: proceed here, because I know he's asking for approval to generate, of course, generations cost money and most of the time the agent will ask for your approval. Sometimes, if you are sure that you want to proceed and you want to the agent to write, rightly, generate the things on the canvas, you can say: don't ask for my approval, generate directly on the canvas, and that will be how the agent will behave. Let's go back to our FLORA.
We can see already our nine shots for the storyboard that are going to be generated and soon we will have six- sorry nine- images and we can move quickly to the production phase. I'm going to go ahead and say to the agents: these set of images are approved and I'm going to state here below the images that I think they are approved to go. I think for now we have nice images here. We have the bomb and the setup looks nice. So the countdown- I guess the pixels- goes down, then the bomb will appear, bomb exploding, and then we have things on to move.
I think overall this is looking nice. The only thing that I will say is that the setup looks different. In this case we have different backgrounds of some with computers, some with just seamless background. So I'm going to go to the agents and I'm going to say: some images had different background. Let's go with seamless black.
So we have- we love- consistence. Once this is done, move quickly to the production phase so we can start creating our videos. Going back to the canvas, we are going to see the agent saying, Okay, all right.,so this one probably needs to be recreated, this one too. And then we have one, two, maybe three images that are going to be recreated and the other ones are going to be just fine. Going back to the agents to see what it says- and a good thing is that always you're going to get the amount of credits or the effort on the agent here.
So it says: swapping the backgrounds to seamless black now. So s3, s4, s5b and s5d, and sc gets a Nano Banana 2 edit at 4K. 4K, which is nice more than to use in this endeavor. Going back to our canvas: you can see here sometimes the agent will move your field, but then you can ask for the agent to make order on this or you can just go ahead and move your two next phases if you want. And why the agent is working.
Usually I'm leaving comments on the images. I'm trying to understand my next step. I'm always trying to be one phase ahead of what the agent is doing. So if I'm trying to retouch or tweak some images here, I will try to understand the next part and how this video is going to look, if this is again accurate enough, if we have something to tweak or if this is enough to get consistent footage. So we have here the seamless black background and the agents already make some order so we can come back actually and move the phases back again.
Perfect, so we have here. And, going back to the agent, we have the new asset. Another good thing is that here on Claude, you will see the asset on the FLORA tab here, so you can see every generation on the agent or on the canvas. Usually I have two monitors, one with the agent, one with FLORA, and then, without switching tabs and without doing anything, just but to leave comments on the canvas and talk to the agent, we can create this footage and these videos. The agent is already had the directive to go to production phase and we're going to get some footage really fast.
We can see here that the storyboard is locked, seamless black, so production is set up and waiting for your go ahead. On the spend, as usual, like we say, in this case we are using Seedance, so it will be 42 seconds of Seedance 2.5 and it will be 51 dollars. We're going to say: go ahead, everything is approved and ready to go and you can see that the agent is very careful when spending is on the, on the medium of your ideas, on the output, which is a really nice thing. Like I said before, if you don't want, this is some is part of my setup and I always try to get a really good understanding of time and effort and money on every project.
I think is a good thing to do when working with this type of video production. This is a short video, so cost is going to be around 60 dollars in total between the images and the video, which is not as much, and it's a quick exploration to pitch something, to get an idea going to lock in a campaign for a client and say that, hey, if you like this, we can certainly go like we did on the second round. We can go deeper, we can get more footage, we can get more consistency, better shots or more control of where we were doing, and increase this budget into 200, 300, whatever is necessary for us.
It's going to use also ElevenLabs for music, which is only 36 cents. Nice one, Let's go back to FLORA here and you can see already that the agent is populating the production phase and getting the prompts ready to animate everything. Also, creating with- oh, both Lyria and ElevenLabs most likely is going to ask me and we will need to choose. So let's just do that and hear the music here. This is nice.
We have two minutes of song here and we can certainly sell it. And again, this part is up to you as human creatives. You can certainly select that. ElevenLabs only did 15 seconds for the video. So, livia, we will need to curate and try to understand which part works better and we don't.
We just have the need to be there. I think that every lap we- okay, we're going to go with ElevenLabs for sure- we are going to skip the curation part on the sound and then start to understand, go back to the model and say: all Seedance clips are generating about five minutes each. I'm also making two music, Lyria and ElevenLabs, so you can click one. So I choose ElevenLabs and find these words. I'm going to showcase something really quickly for us.
Since we are on a live demo and we have limited time, I don't want this to go without showing, so most of the times we do this type of presentations, they ask us: hey guys, how did you do those animations, those motion graphics with images? So here we have inside the presentation. We have a web app for you, so you will get the link after this webinar is ended and then you will be able to use this app for yourself. As you can see here, this is an app called Bloom Studio and basically what you have is presets animations for your images. It's super easy and intuitive to use and you can even move the here the axis to get the best perspective for your photos.
You can increase the card size, you can make them a little bit more spread if you want, change the aspect ratio, do a lot of things with these animations, and this is free for you to use. You can export a loop and then put this loop in your presentations as well. We have a review ring here, a library sphere like the ones that you saw on the cover. We have a spiral stream. This one is one of my favorites campaign stack also.
Remember that once you change the axis here, you can do a lot of things with this. You have another. One of my favorites are brainstorming animation here and several things to change. Of course. You can change the backgrounds, you can change the like we were saying, the aspect ratio of it, the tilt of the of the animation, the corner radius of the images if you want to make them circle.
There is a lot of things here. You can shuffle your images as well if you don't like fully like what you're seeing. But adding your images too will be a really nice, I think, to do. Let's go with, let's go raise animation grid. We have a card channel too.
This one is really nice cover flow for your heroes. We have the bloom and totem wall. Finally, the orbit bloom- another one classic animation on after effects. Now you have this for free here to be able to use this on your presentations. So everything that we just saw on these agentic workflows on FLORA is the way that we are working on highly intelligent systems where the human input is still one of the most important parts of the whole process.
Your ideas, your checkpoints, the quality that you have, the expertise that you have be behind every, every move that you do on this type of campaigns. But this is really helpful for any campaign concept team: to pitch anything to clients super fast, to make social content as scale. Remember that here you can have dozens of these videos in one day, which was totally impossible before, before AI. Any pitch deck that you will need to create, move boarding or anything, will be super easy to do with FLORA and agents. You can use Claude, or you can use certificate or any agent that you see fit for this task.
And finally, to recap, three things to remember. We have one canvas beating six tools every time. Having everything together on the canvas, making every phase of the project there, makes us more efficient and grants us the speed that we need to deliver as fast as we can, keeping the quality and keeping a high standard on everything we do. The volume changes the work. In this case, again, we can generate dozens of directions.
We can choose, we can curate. The content is not just one, as we were used to do before on this industry. And finally, the pipeline doesn't end at the storyboard, which is something great to me. We can quickly showcase some motion, some videos, and say, hey, this is what we can do. Give me three days and we can make this amazing.
Now you can leverage this technology to translate those thoughts into things that are really, really amazing. So I agree, it's the best time to be a creative in the history of creative. So just embrace that and get your idea going.
