How does the built-in credit system work?

Every account receives a free credit allowance that renews each month on your signup date — the same day-of-month you joined. Credits are consumed atomically only when a hosted cloud model executes successfully. If a model fails or returns an error, the credits are immediately refunded. You can view your real-time balance in your project toolbar.

Can I bring my own API keys?

Absolutely! Under your Account Settings, you can configure your own Google AI (Gemini/Imagen/Veo) or xAI (Grok) API keys. When you do, we route executions directly through your credentials. In this case, no credits are deducted from your balance, allowing you to pay raw API costs directly to the providers.

Are my API keys stored securely?

Yes, security is a core pillar of genvi. Your API keys are encrypted with authenticated AES-256-GCM and used only to execute the requests you run against the AI providers you choose. They are never sold and never logged in plaintext, and they are removed when you revoke them or delete your account.

What is Chat Studio, and when should I use it instead of a workflow?

Chat Studio is the no-nodes way to create. You describe what you want in plain words and get an image, video, music bed, voice-over, copy or 3D asset back, then keep replying to refine it. It reaches the same model catalog and the same credit system as every other studio, and saves everything to your File Repository automatically. Use chat when you want one asset and expect to iterate conversationally. Use a Node Workflow when you need the same pipeline run repeatedly, or several models chained in one pass.

How do I keep the same character or product consistent across generations?

Save it once to your Reference Library. A reference is a name, a description written for a model to read, and up to twelve pictures. Type @ followed by its handle in any prompt box, whether that is Chat Studio, Node Studio, the whiteboard, either composer, 3D Studio or Genvi Agent, and the same description and the same reference images reach the model every time. Your wording is passed through verbatim rather than paraphrased, because a rewrite of the detail you care about is exactly what drifts between shots. Group several references into a set and one mention brings all of them, and when a generation comes out right you can save it straight back to the library from wherever it appeared.

What is the difference between Workflows and the Whiteboard Canvas?

Workflows are structured node-based pipelines representing DAG execution trees (topological routing of AI generations from inputs to outputs). The Whiteboard Canvas is an infinite, unstructured sandbox where you can freely draft flowcharts, mind maps, sticky notes, Kanban boards, and freehand pen drawings. It also features a natural-language AI chat assistant to generate diagrams instantly.

Who owns the commercial rights to generated media?

You own 100% of all generated and compiled assets. All images, layouts, sequences, character sheets, sprite grids, and exported videos are fully yours to use commercially without any licensing fees, royalties, or attribution required.

How does the Video Composer render browser-side?

The Video Composer uses WebAssembly-based FFmpeg (FFmpeg WASM) directly inside your browser. This enables offline-capable, high-performance rendering that operates entirely locally on your device without transmitting your heavy video clips to any server, saving you render fees and maintaining absolute file privacy.

August 22, 2026

Tell us when a result misses

Every generated image, video, sound and model now has a thumbs up and a thumbs down, with room to say what went wrong.

We carry more than a hundred models. Deciding which ones deserve to be the default is the hardest ongoing call we make, and honestly, we're not the ones best placed to make it. You are. You're the one who looked at the output.

What changed

Generated results now carry a rating control wherever they show up: in Chat Studio, on a node in the canvas, and in 3D Studio. Thumbs down opens a comment box. Nothing is required, and nothing blocks what you were doing.

What we actually do with it

Every rating is tied to the exact model and prompt that produced the result. That turns "the images got worse lately" into a specific model we can go and look at.
When a model is quietly failing one kind of prompt, it shows up as a cluster. That's usually what decides whether it keeps its place on Auto.
Written comments go straight to the people who tune the catalog. They're the most useful thing you can send us, by a distance.

A rating isn't a support ticket. If something is broken and not just disappointing, come and find us on Discord. Nobody will reply to a thumbs down.

If you've got thirty seconds after a generation that let you down, that's the most useful thing you can do for the catalog.

Tell us when a result misses

What changed

What we actually do with it

Product

Company

Legal