Tetiana Stepanets: Why I'm Building Lilidi AI — LiliDi Blog
Meet Tetiana Stepanets, founder and CEO of Lilidi.ai, and the story behind a multi-model generative AI platform for image, video, audio and creative workflows.
By Tetiana Stepanets
From bootstrapped e commerce and product building to a unified generative AI workspace — the founder story behind Lilidi.ai. Tetiana Stepanets — Founder & CEO, Lilidi.ai — Brussels, Belgium — Updated July 2026. At a glance Tetiana Stepanets is the Founder & CEO of Lilidi.ai , a generative AI creative platform that brings leading image, video, audio and creative AI models into one workspace. Before Lilidi, she spent years as an operator — building direct to consumer e commerce brands, running a coffee shop business that grew into a franchise across Ukraine and Poland, and shipping software products. Her previous entrepreneurial experience includes Crema Caffe , Homsto (a printable vinyl sticker paper brand sold through Amazon), and SeaOcean , an AI powered SEO and GEO platform she co founded. Today, she is focused on building Lilidi.ai as a simpler, more transparent way for creators,
marketers, founders and brands to work with generative AI. Founder & CEO · Generative AI · Creative AI · DTC & E commerce · Product Development · Brussels, Belgium A founder story, not a pitch I didn't start Lilidi because the world needed another AI wrapper. I started it because I became the user I couldn't find a good product for. Every day, across half a dozen tabs, I was assembling a creative workflow out of tools that weren't designed to talk to each other. One subscription for images. Another for video. Another for voice. Another for editing. Another for upscaling. Every tool had its own billing, its own asset library, its own idea of what "generation" meant. The technology was extraordinary. The workflow was exhausting. Lilidi is the answer to that exhaustion. It's not a rethink of AI. It's a rethink of how a working creator actually sits down to use AI. Before AI, I was an
operator For most of the last decade I wasn't in AI at all. I was in the trenches of small business. I co founded Crema Caffe , a specialty coffee shop brand that grew from a single location into a franchise network across Ukraine and Poland. Coffee taught me operations — margins measured in cents, staffing, supply chains, brand consistency across dozens of locations, and the difference between a customer and a regular. Then I built Homsto , a direct to consumer brand selling printable vinyl sticker paper on Amazon. Completely bootstrapped. No investors, no runway, no soft landing. When you sell physical product on Amazon you learn very quickly what marketing actually is: it's a photo, a headline, a video, five bullet points and a review score. Every listing is a tiny landing page fighting for attention. Every A/B test costs real money. More recently I co founded SeaOcean , an AI powered
SEO and GEO platform that helps sites get discovered by both traditional search engines and the new generation of AI answer engines. That was my first close up look at how large language models are quietly rewriting how people find information — and how brands have to write for both audiences at once. None of that background is glamorous, and that's the point. Building real businesses teaches you that creative production isn't an abstract intellectual exercise. It's an assembly line. You need product photography, ads, hero visuals, thumbnails, short form video, voiceovers, music beds, social posts, email creative and about twenty variations of all of the above — every week, forever. Speed matters. Cost matters. Consistency matters. That is the operator lens I brought to Lilidi. Then generative AI changed the creative stack Somewhere between 2022 and 2024 the tools we'd been waiting for
finally arrived. Image models became genuinely useful. Video models crossed the threshold from "novelty" to "production." Voice cloning and text to speech stopped sounding robotic. Music generation became listenable. Editing models — inpainting, background removal, upscaling, style transfer — became reliable enough to trust in real work. The technology curve went vertical. And then something strange happened: the user experience got worse. Every capability shipped as its own product. Each with a different UI, a different pricing model, a different credit currency, a different asset library. If you wanted to make a single 15 second ad — a still image, a short animation, a voiceover, a music bed, one round of edits — you were logging in to five different platforms, paying five separate subscriptions, and manually shuffling files between them. Powerful models, fragmented workflow. That gap
is the entire opportunity. The frustration that started Lilidi The problem I set out to solve wasn't a shortage of AI models. It was access, orchestration and honesty about what happens under the hood. Creators don't think in tasks. We don't sit down and say, "Now I am performing an image generation task." We think in ideas. An idea starts as an image. Then it wants to be a short video. Then it wants a voice on top. Then a music bed. Then a variation, and a variation of the variation. The workflow should follow the idea. Instead, the idea has to bend around whichever product happens to own each capability. I wanted a place where the prompt bar didn't care what modality the output was in. Where the credit balance didn't reset depending on which model I picked. Where the model choice was visible, not hidden behind a friendly sounding "Auto" that quietly routes you to whatever's cheapest
for the provider. And where the work I made was mine — searchable, remixable, downloadable, portable. That place didn't exist. So we're building it. What Lilidi.ai is Lilidi.ai is a generative AI creative platform that gives creators and brands access to leading image, video, audio and creative AI models through one workspace, one credit balance and a unified workflow. The philosophy is simple: One prompt bar for every modality — image, video, audio, edit, upscale, remove background, vectorize, poster and more. One credit balance across every model. No juggling five separate subscriptions. Transparent model choice. You see which model ran your prompt. When you want control, you get it. When you want convenience, "Auto" tells you what it picked. Your work is yours. Every generation is saved, searchable, renameable, downloadable and yours to remix. Move between modalities without
rebuilding your workflow. Start with an image, animate it, add a voiceover, drop in music, ship the whole thing. You can browse the current model lineup on the Models page — it grows almost every week as new flagship models ship. Why one AI model is not enough There is no universal "best AI model." Anyone selling you that story is selling you their inventory. Different models are genuinely better at different things: Some models excel at photorealism and product photography ; others at stylized illustration and brand art . Some are unmatched at typography and posters ; others handle cinematic video or fast rough iterations . Voice models have distinct timbres, accents and delivery styles that suit different jobs — a warm brand narration is not the same voice as an energetic 5 second ad tag. Music models split cleanly between cinematic score , radio ready song and short loops for social .
Editing and enhancement — background removal, inpainting, upscaling, style transfer — is a completely separate skill from generation, and the best model for each task is rarely the same. The future creative stack is multi model. Not by ideology, but because the alternative — pretending one model wins everything — leads to worse work. Lilidi is being built around this reality. The future of creative AI isn't one model winning everything. It's being able to use the right model for the right idea without breaking your creative flow. — Tetiana Stepanets , Founder & CEO, Lilidi.ai Building for creators — not around lock in The product principles we hold ourselves to are unglamorous but load bearing: One prompt bar. Multiple models. One credit balance. Transparent per generation cost, shown before you spend. No subscription maze, no artificial gates around which model you can pick. No hidden
routing when you want model control. Outputs stay with you. Rename, organize, download, remix, share. Move fluidly between image, video, audio and creative editing tools without rebuilding your workflow every time. The moment any of these principles starts to bend under commercial pressure, the product stops being the thing I wanted to use in the first place. What I've learned building Lilidi Small teams get to make big decisions quickly, and pay for them quickly too. The thing I didn't fully appreciate when I started is how much of the work is not the model itself. Wiring up dozens of upstream providers, keeping their APIs healthy as they change under you, orchestrating queues, handling failures gracefully, showing accurate costs before a generation runs, persisting outputs, making the whole thing feel fast — that is the actual product. The models are the easy part. The other thing I've
learned: user feedback beats founder intuition, but only when you know how to listen to it. Users don't ask for "multi model orchestration." They ask, "Why did this take so long?" and, "Why is my video black?" and, "Where did my last thing go?" Every one of those is the same underlying problem in a different disguise: friction in the workflow. My job is to translate the complaints into architecture. What I'm building toward The long term direction for Lilidi is not "aggregate more AI models." Aggregation is the floor, not the ceiling. What I'm building toward is a complete creative environment where a working creator can move between models and modalities without friction — where you can bring an idea in the door as text and leave with an image, a video, a voiceover, a track and a poster, without ever thinking about which vendor is behind which button. That means: Broader access to
leading AI models as they ship, without waiting for the next quarterly release. Image, video, voice, music and creative editing living in the same environment, with the same asset library and the same credit balance. Stronger multi model workflows — hand offs between models that feel like one continuous session. A community layer that rewards remixing and gives credit back to the original author. API and MCP integrations for advanced users who want to wire Lilidi into their own stack. Transparent model choice and pricing at every step. Always. FAQ Who is Tetiana Stepanets Tetiana Stepanets is the founder and CEO of Lilidi.ai, a generative AI creative platform. Before Lilidi she built and operated businesses across direct to consumer e commerce, hospitality and software — including Crema Caffe, Homsto and SeaOcean. She is based in Brussels, Belgium. What is Lilidi.ai Lilidi.ai is a