How a Figma Weave Insider Gets Creative Teams to One Click

Skeptic to believer in one demo.

Episode 88 has a guest with a title Drew had never heard before. Ron Baranov is a creative workflow advocate at Figma Weave, and he spends his days inside client teams.

“Ron has a title that I hadn't heard before, but makes perfect sense in where we are today, which is Creative Workflow Advocate.”

Ron spent ten years in video production. Seven were freelance. Three were in-house at an Israeli venture firm, where he built its video studio from scratch.

“I wasn't the best director. I wasn't the best editor. I wasn't the best cinematographer. I wasn't the best color grader. … But basically I knew how to do all of those stuff in a sufficient degree.”

Then a portfolio company hosted a small demo for a startup called Weavy. One of the founders was a friend from Ron's teenage years, so he took a day off to see it.

“There's pictures of me from becoming skeptical to becoming a believer on the same event.”

“There's a really great way to control AI for your creative outputs. And you can also integrate LLMs in order to do that.”

He went home and told his wife he would never leave his comfortable job. If he ever did, it would be for a company like that one. From then on, he was hooked.

“I was thinking about workflows before I go to sleep.”

He started posting his own workflows. The team noticed, and Ron joined as the first hire after Figma's acquisition of Weavy.

“I'm the first employee after the acquisition. … This is how I brand myself.”

Rory remembers why it landed when it did. Every tool lived in its own silo, and creators argued about which one to pick.

“Everything was so siloed. Every tool was in its own little space, and you had to have 50 subscriptions to get anything done.”

So what does a creative workflow advocate do all day? First, Ron says, you have to be the expert on the tool. Then you wear four hats.

The four hats

• 01

  Content and education

  New models, features and workflows, explained on YouTube and in brand material.

  The node basics he runs for every new team. Chapter 4: Basics →

• 02

  Enterprise onboarding

  A team of seven advocates takes client teams from zero to building their own workflows and tools.

  A furniture maker's wood textures and a loop-animation tool. Chapter 2: Clients →

• 03

  Product feedback

  Super users who report to product and engineering, honestly, with no sales quota.

  Gen Effect, the feature he thanks the engineers for. Chapter 10: Gen Effect →

• 04

  Workshops and stages

  Workshops, conventions and talks around the world.

  A two-hour live build in Milan, and Rory's Config talk. Chapter 12: Literacy →

“We only think about the product. We are not sales or marketing.”

That last hat matters to him. When something in the product isn't good yet, the advocates can say so, to the product team and to the client.

The gap is the LLM.

Drew spent a decade in B2B marketing. The best thing he could do was get on customer calls.

“You hear people describe your product in a way that maybe you hadn't thought of before.”

So he asks Ron whether clients know what they need. Ron says it depends on who is on the team.

“When you have a creative champion on a team, this is amazing. Because usually it will be the guy that will become a builder for that team.”

Most teams still need the advocate. Ron sorts the work into two pillars. Pillar one fixes a process the client already knows is broken. His example is a large furniture company with a library of wood textures.

“I'm sitting in a meeting speaking about textures of wood. For me, it's like, my God, this is great.”

The ask: pick a new wood texture and apply it to existing furniture photos. Pillar two is the work nobody thought to ask for, because it wasn't possible before.

“One of the greatest workflows we created for all of our enterprise clients as a use case, just take your stock imagery and create videos out of it.”

Drew asks what Ron tries to learn first in discovery. The answer is the same across his enterprise clients.

“The thing that will be lacking the most in my opinion, and this is throughout enterprise clients, is the understanding of what LLMs can do for them.”

These creative teams are top professionals, and their single outputs look great. Ron's job is to show how an LLM can carry that output through the workflow they already built.

“LLMs that will write the prompts for you. And we'll do it in scale.”

Same job, new input.

Rory has watched teams start building. The first thing they do is overcomplicate it.

“There's a tendency to over complicate workflows, right? Like, I'm sure you've gotten a lot of clients who come in there and there's like 17 branches of the workflow.”

“If you're building for like tools and teams, right, like it has to be a certain way.”

Ron defends the one-off. Weave's brand team makes the company's brand videos in it, and each of those workflows gets used once.

“I'm so jealous of your one-off workflows.”

Enterprise work runs on a different rhythm.

“When we are working with enterprise clients, most of the time, the nature of their work will be we need to do the same things with a different input.”

So how do you move someone from one-offs to tools? Ron asks one question.

“Most of the time when you will ask them, is this something that you are doing repeatedly? They will answer yes.”

One-off or tool?

Ron's test is one question: is this something you do repeatedly? Call each job, then see where it goes.

1. 1/7

   A launch video for a new plan

   Episode: Build a tool

   Made once, shown once. Ron is jealous of these. Build it on the canvas and let the workflow sit.

   Stays on the canvas

2. 2/7

   Loop animations for each new character

   Episode: One-off

   Same job, new input every time. This is the white-label loop workflow enterprise clients run.

   Tool mode · one image, one word

3. 3/7

   Branded invites for ten event speakers

   Episode: One-off

   Ten portraits, one look. Ron's invitation machine lands each one in its Figma frame.

   Figma node · one frame per speaker

4. 4/7

   A mood board for one pitch

   Episode: Build a tool

   One pitch, one direction. Explore freely and move on.

   Stays on the canvas

5. 5/7

   A new wood texture across a furniture catalog

   Episode: One-off

   Pillar one: a process the client already knows. Pick the texture, apply it to every existing photo.

   Tool · texture in, catalog out

6. 6/7

   Product cutouts for a fashion catalog

   Episode: One-off

   Extract the product, replace the background, repeat. Ron runs this as a Weave tool inside Figma.

   Weave tool in Figma

7. 7/7

   A personal side project, just for fun

   Episode: Build a tool

   Make it because you want to. Rory says personal builds are where new ideas come from, and Ron's casting game started as one.

   Stays on the canvas · maybe a tool later

From there, teams learn to think in LLMs, iterators, prompt variables and tools. Drew learned the way most people do. He opened other people's workflows and worked backward.

“Just take a workflow link you see out there, open it, duplicate it into your files and just see what is inside.”

Six models, one prompt.

Ron runs the intro he gives every new client. One plain prompt, a man lying on the grass seen from above, goes into one image model.

“It's super generic. Okay. It's not tasteful and it feels like AI.”

Then the same idea with an intentional prompt, sent to six image models at once. Each model reads it differently, and that's where taste comes in.

“You can see that different models perform differently in the same context. So it's the same prompt and we're getting six different results. And now I can apply my taste on the thing that looks best for this specific project.”

He picks the GPT Image 2 result, changes the outfit with image to image and checks the change with a compare node. The image becomes a video across six video models. Then he pulls a frame from the video and crops the original, to build context for the next step.

“One of my favorite nodes, by the way. I love that compare node.”

Next comes the LLM. Ron opens every class by asking it for a pizza recipe, so people see it's the same chatbot they use all day. Then he gives it a creative job.

“Inside of Weave, you want to give it creative tasks. You don't want to ask it to plan a trip to Italy or give you a pizza recipe.”

His instruction: write four image-to-image prompts for new camera angles of the scene in image one, using the character in image two, and change nothing but the camera.

“I didn't write the prompts. I only gave it the creative direction role. And I told it, what are the inputs? What are the outputs? I did a contract with it.”

Four prompts, or one contract

Ron wants four new camera angles of one scene. Flip the switch to compare the two ways to get them.

Write each promptBy hand

Prompts you wrote 04 04 08 08 Angles back 04 04 08 08

You write

1. Prompt 1 · low angle, same scene, same character
2. Prompt 2 · over the shoulder, same scene
3. Prompt 3 · wide shot from the tree line
4. Prompt 4 · fisheye, close to the ground

Four prompts, four angles. A new scene means four more prompts.

New scene arrives: Write four more prompts for the new scene.

Write one contractRon's way

Prompts you wrote 01 01 01 01 Angles back 04 04 08 08

You write one system prompt. The workflow does the rest.

1. In · image 1, the scene
2. In · image 2, the character
3. LLM · write 4 prompts, change only the camera angle
4. Array and list · split the 4 and run them
5. Out · 4 new angles of the same scene

One contract, four angles. A new scene means zero new prompts.

New scene arrives: Swap the input image and run it again.

An array splits the answer into four prompts and a list node runs them. Four new angles come back, and Ron keeps the fisheye. Drew sees the mindset shift that in-house teams still miss.

“The mindset in 2026 … is you're having LLMs write everything for you, right? Like if you give it enough context, it knows how to speak to itself better than you can speak to it.”

“If you're getting somebody on this, you got to get them activated right away. The LLM writing the prompt is a big thing.”

Ron shared the intro workflow from this demo. Source: Intro to Figma Weave workflow (Ron).

Know your room.

Drew raises the question every enterprise buyer asks: is this model approved?

“Is there any concern or questions you get about, is this model approved? Can we use this model?”

“Giving them complete control, complete control over the models that they could allow their users to use.”

Weave has a model management dashboard and a list of verified models. Ron says the account team works it out with each client's legal team, and some legal teams draw a tight line.

“We do see some enterprise clients working only with four image models because this is what their legal team is allowing them.”

Rory turns to the demo itself. Anyone who trains teams has to show the tools without drowning the room. Ron built this one with his team to cover the basics before any use case.

“First I'm asking what is the timeframe I have and what is the goal of this session.”

His early demos were rough, and the feedback shaped the next ones. The pizza recipe exists because people didn't see that an LLM node is the chatbot they already use. Next, Ron was heading to Milan for a two-hour live workshop.

“So iteration, it's the answer. Iteration being thoughtful and minded to want it to be better and also timeframe and goals.”

“I will show the usability more than the capability.”

Rory thinks the teaching now matters as much as the build.

“The education portion is just as important as doing the actual work at this point, because you have to number one, you have to sell it to whoever is buying it, not just the tool, but the capability of it and the trust that this stuff can be done.”

“So you have to be able to explain it in a way and actually make it replicable for people because if not, they just see it as a cheap trick.”

Ron's goal for the episode is smaller. Spark the curiosity, and the learning follows.

“If you will put your mind into that software, you will be curious and you will want to learn.”

One builder, one input.

Drew asks for the boring workflow that saves teams the most time. Ron opens a white-label loop animation, a workflow some enterprise clients run.

“This is a real workflow that some enterprise clients are using. And basically what they want to achieve is a loop animation. That's it.”

Four animals in one style get four actions: jump, spin, kick, wiggle. The client prompted Seedance 2.5 by hand and did it well. The advocate's move is to put an LLM in front of it.

“This is exactly the example of seeing an enterprise client do it once well. It knows how to do it well. But then as an advocate, we can tell him, hey, we can automate this process.”

The system prompt tells Seedance to use the same image as the first and last frame, so every clip ends where it began. Ron didn't write that long system prompt alone either. He worked it out in a conversation with an LLM, and the first versions didn't work.

One word in, one loop out

The white-label loop tool, run by someone who never opens the nodes. Hit run and watch it fill.

Loop tool · character 1 of 4

Input

one character image

Style

locked by the builder

Request

jump

Length

5 seconds

Steps run 06 06 Prompts you wrote 00 00

1. 01 Character image and request go in Input Step 1
2. 02 LLM reads the builder's system prompt LLM Step 2
3. 03 LLM writes one Seedance prompt for the action LLM Step 3
4. 04 Same image set as first and last frame Rule Step 4
5. 05 Seedance 2.5 renders five seconds Video Step 5
6. 06 Clip ends on the start pose: a loop Output Done

Then the workflow goes into tool mode. The builder shares it, and the rest of the team only touches the inputs.

“You don't need everybody to be able to build these things. You need a few people or one person even that could build this. And then you've got everybody activated once you've got that.”

“The simpler these things are, the more repeatable they are.”

Ron's app-mode example is a style guide maker built by a fellow advocate, one of Weave's most popular workflows. The illustration style is locked, two LLMs do the describing, and the only input is a new theme. His test was a dancing honey badger.

“Imagine that this is a machine that the marketing team can use for social outputs without having to speak with the creative team all of the time.”

“The workflow could be very complex, but the app only one input.”

Live in the frame.

Rory asks about the Figma node. It shipped after the hosts' own Weave episode, and he wanted to see it.

Before the node, you made assets in Weave, downloaded them and imported them into Figma. Now you copy a Figma frame and paste it into the Weave canvas. Ron pastes in a frame and connects a new image to it.

“Now this guy is here. Okay, he's here. He's connected to my frame.”

Rory explains it for listeners. A Figma frame has layers, like any design file, and the node can target one of them.

“You're able to only control the one layer of the frame if you want to.”

Ron's real example is an invitation machine. Drop in a photo of any speaker, a name, an event, an outfit and a color palette. An LLM writes the prompt that restyles the portrait, and the result lands in the Figma frame.

“So imagine that I have 10 different speakers. They are all going to speak in the same event. And they want to do a branded invitations of those speakers.”

Rory asks whether you can flip between versions inside Figma. For now, the linked frame follows the canvas. When you like a result, you copy and paste the frame to keep it.

“Once you're satisfied with the result of a Weave frame, you can just copy and paste it.”

Hours behind one click.

Rory calls himself Rory MCP. He has Claude skills that call Weave tools he built, like a quick background removal, and he uses Claude inside his workflow testing.

“It's great to have like, Claude as a partner in there, not only just helping like with the workflow and sort of troubleshooting, but also like version control.”

“You're making a big change to this, you should just duplicate the workflow, put it in a folder.”

He also catches Claude reaching for a prompt when a visual reference would work better. When it offers to build him custom nodes, he tells it not to complicate things.

Ron picks up the thread on system prompts. The pitch out there is a full brand book in one click. He tells workshop rooms what sits behind it.

“Behind that there are hours of hours of clicking and trying and iterating and not understanding what's not working and trying to solve it. … It's the new part of the job.”

“Usually the last 10 % of the things that are important to you will take a while to try and figure out when you run it in scale.”

“Getting to that 85 to 90, and then it's really like that 10 to 15% takes just as long, if not longer.”

Rory is living it with a client's product shots. The workflow is simple, but the size, position and background color must match on every image. The MCP lets Claude see the outputs.

“After a certain point, you're just downloading and downloading and downloading and dumping it in there and hoping it understands the context versus knowing the context and not having to decipher it.”

The loop behind one click

Ron's point at 58:02: a one-click tool hides hours of clicks. This is the loop those hours go into.

ONE CLICK

1. 01Build it
2. 02Run at scale
3. 03Spot the miss
4. 04Fix the prompt
5. 05Save a version
6. 06Ship one click

Where the hours go

• Hours of clicking behind every one-click tool (Ron)
• The last 10% takes as long as the first 90 (Drew)
• Claude reads the outputs through the MCP (Rory)
• Version history keeps the last good prompt (Ron)
• Out: one click a teammate can trust

Ron closes the loop inside Figma. Weave tools sit in the Figma panel, so a designer never leaves the file. He runs Extract a product to lift a camera out of a photo, then Replace Background.

“What I like about it is the fact that you can create your own specific tool for your own specific needs.”

“You can build your own tools in Weave, publish it to the community, and just work with it here.”

Ron shared the Extract a product tool. Source: Extract a product from an image, Figma Weave tool (Ron).

Make it a game.

Ron asks to show something personal, a workflow nobody asked him for. Rory and Drew agree those are the best ones.

“I'm a bit lazy about the inputs I want to input, but I do want my LLMs to give me a lot of, like to do it almost like a game, okay?”

There are five inputs: gender, age, ethnicity, a free-text note, and a world and era. His test is a young filmmaker in 90s New York street fashion. Everything goes into a prompt variable and on to two LLMs.

“This LLM is a casting director that will give me, I think 10, yeah, 10, profile images of new characters.”

The second LLM returns 16 haircuts, 16 outfits and 16 accessories on contact sheets. You crop the ones you like, and the workflow builds character sheets. Rory stops him to point out the trick.

“If you're doing something from the 90s, and you don't know how to describe the haircut or know what you want, you're endlessly generating. This to me is such a nice fix. It's just like show me 16 versions of it.”

Then Ron opens version history on a prompt node, and Rory learns it exists.

“Hold on, hold on. Has that been there the whole time?”

Ron's stress test for the LLMs was a teenage fluffy monster in pastel colors. The grid held, and the casting director handled a creature it was never told about. Drew loves the layout.

“It feels like you've got this big desk and you've laid it out. You've got everything on the table that you can glance at, pick from, choose all on one.”

Ron's first workflow was a B-roll generator with so many arrays he can't look at it now. Newer models do most of that job in two nodes. So the question he asks is where to keep the human.

“What are the friction points that are important to me as a creative that I want to put my taste in?”

“The parts we keep for ourselves and most of the time it changes, but it will be the creative direction and the creative choice and the places where you need to apply taste.”

Rory takes the idea home. He might run his 3D light sphere references the same way, sixteen at a time.

“This is why I tell people all the time to do stuff, basically for personal use, you did this because you want to do this, right? And you're thinking about it a completely different way than this is for a client.”

Rory asks whether the advocates get competitive. Ron calls it professional. The team's passion is what inspires him.

Ron shared the character workflow from this chapter. Source: Create character workflow (Ron).

Build your own node.

Drew asks where node-based tools go next. Ron calls himself a true believer in building over prompting.

“When you don't get it or when you want to change something, I rather to build it from scratch. The same prompt you've built, I rather to build it and to understand what I'm going to do with anything that is not going to work that I want to change.”

His pick for what comes next is Gen Effect. You describe an effect in plain words, and the node codes it on the canvas.

“And then I can generate it and I'm basically vibe coding an effect on the canvas. And this is going to be my personal node.”

He asked for a film look creator with grain that leaves the video untouched at zero. It came back with grain and faded blacks. Two words, surveillance camera, came back with a green tint control he never asked for. Then he added a pin transform and a glass reflection.

“But look how smart it is without me asking.”

Drew likes that the choices stay with the creative.

“There are still creative decisions that you want to make that make you feel like you're a part of the process, right?”

Save the node and it's yours for next time. Rory used Gen Effect for torn paper edges on a collage. When the result has to be perfect, he adds one more step outside the node.

Ron shared the Gen Effect workflow. Source: Gen Effect workflow (Ron).

Find the junkie.

Ron flips the interview. What do Drew and Rory see when they bring AI into companies?

“If you're trying to be serious about operationalizing and scaling this process, you do need templatization. You need a system, right?”

“The ability to package things in a singular way, get one approval versus five or six, and the ability to kind of like activate people as quickly as possible.”

Rory's first job inside any company is a person.

“But me, on my personal side, I have to find the person. I have to find the junkie.”

“Who's the person that is going to just like take this and run with it? Because I can't be here all the time.”

Drew names the hard part. Even when people see the value, each one decides alone whether to go further.

“Change management adoption is the hardest part.”

“It's cool if I can show you and deliver this stuff, but if you don't do anything with it after, that makes me feel a little disappointed in maybe how I approached it.”

Ron's advice comes in two parts. The first is about what you leave behind.

“We are trying to teach you how to fish, but also give you some fish.”

Rory adds that you learn the most in person: which tools a team uses and how the office runs. Drew adds the split between the person who buys and the person who uses.

“Who's making the decision versus who's using it, right?”

Ron's second part is for managers.

“In every team you can see the people who are leaning more to want to adapt or to be more curious and it could be very surprising who that person is on your team.”

“You can tell right away there's someone that you can see the little kid sparks up inside of him.”

The new literacy.

Drew's last question: what should creatives lean into now? Ron sees the same gap at every big client.

“I see a gap of a person who will be the owner of, can call it creative technologies. We can call it creative workflow advocate. … There are so many titles for this job.”

“More than just a VFX artist, for example, or a 3D artist. You should expand.”

The second gap is the LLM, and Ron wants creatives to see it as more than automation.

“It could also be like the things you saw here, creating creative tools that will assist you and your team to expand your ideation.”

He warns clients the work has a dull middle, and that the middle is the job.

“It's not going to be the most fun thing you're going to see in your life. It's going to be a bit mundane for some parts.”

“What am I trying to achieve? What are the inputs that I have? Right. What are the LLMs that I'm going to put in between”

“I think we say that the creative workflow building is the new literacy.”

Rory backs it with a story. Figma paired him with Ron to prepare his Config talk. His first draft for a twenty-minute slot was a lot of slides.

“My presentation was twenty minutes. I think I delivered the first draft to Ron and it was me a hundred and thirty seven slides.”

“But maybe we should take 35 minutes.”

Ron tells listeners the talk is up on Figma's YouTube channel. Before the wrap, Rory asks about the new office hours.

“Every Thursday on 4 PM European Central Time, we are doing an office hour session with a topic in mind.”

The sign-off rule stands: you don't have to like the video, but you have to tell somebody. This week's shout-outs go to handlers, fixers, baristas and engineers. Ron adds one more group.

“You can call us Weavers if you want.”

New to Weave? Ron points beginners to the first-steps tutorial. Source: Weave first steps, tutorial on YouTube.