Today Sumi Studio opens its doors. What started as a simple conviction — that AI filmmaking should feel like directing a film, not like ferrying files between a dozen disconnected tools — is now a working studio: an infinite canvas where scripts, storyboards, shots and edits live side by side, and every one of them can actually run.
The film in this post's header was generated end-to-end inside Sumi Studio, like every other showcase asset on this site. That is the standard we hold ourselves to from day one: if we show it, the canvas made it.
Why an infinite canvas
Ask any director how a film actually comes together and you will hear about walls. Walls of index cards, printed frames, scribbled arrows, alternate endings taped over the endings they replaced. Creative work is spatial — you think by arranging things next to each other. Yet most AI tools compress that work into a chat box or a render queue: prompt in, clip out, context gone.
An infinite canvas restores the wall. On Sumi Studio, a project is a space you move around in, not a form you fill out. Your script sits next to the storyboard it produced; each storyboard frame sits next to the shots generated from it; rough takes sit next to the versions that replaced them. Nothing disappears into a history tab, and when a scene refuses to work you can usually see why — because the whole argument of the film is laid out in front of you.
A project is a space you move around in, not a form you fill out.
And unlike a cork board, every node on this wall is live. Re-run it with a new prompt. Branch it into three variations. Feed its output into the next stage. The board is the pipeline.
Script to screen on one surface
The workflow reads like a film credits list, in order. Type a logline and an agent drafts the script. Break the script into a storyboard with a click. Turn any frame into motion with the strongest video models in the industry. Then cut, upscale and export — without ever leaving the surface where you started. Each stage hands its output to the next as a node you can inspect, override or regenerate, because a tool that hides its intermediate steps is a tool you cannot direct.
That word — direct — is the whole point. Generative models are powerful, but power without control produces beautiful accidents. Everything in Sumi Studio is built to give you the second take: pin what works, vary what doesn't, and keep the full history of both on the board.
One model layer, every model
No single model wins every shot. Seedance renders cinematic realism at native 4K; HappyHorse excels at expressive, physical motion; GPT Image and Gemini handle design frames and style exploration. Rather than crowning one engine, Sumi Studio ships a unified model layer: you pick the right tool per shot the way a cinematographer picks a lens, while your assets, credits and version history stay in one place.
This matters beyond convenience. Models are improving month by month, and a studio built around any single one of them inherits its ceiling. A unified layer means new models slot into your existing workflow as they arrive — your storyboard does not care which engine renders it.
Launch Season Creator Awards
To mark the launch we are running the Launch Season Creator Awards, sponsored by Alibaba Cloud — the compute behind this launch season. For 30 days from launch day, anything you create on the canvas and publish to the Community is eligible for the credit prize pool, and standout work gets featured on the homepage. The full schedule, entry rules and reward tiers are on the contest page — it is linked right from this blog.
An open invitation
This is day one, and day one is honest: there are rough edges, and the roadmap is long. What we can promise is the direction — one surface, every stage of the craft, and real outputs you can publish, enter into contests, or cut into something bigger. Open a canvas, drop in the idea you have been carrying around, and see how far it can go.