AI Video & Animation:
A Small Studio's Guide

The short answers

• You do not need an enterprise agreement to protect your work on Google Veo (paid tier) or Kling (paid API). Both already promise not to train on your uploads at normal paid pricing — no lawyers, no minimums.
• You do need to think about enterprise (or avoid the platform) with Runway. Its standard and even its paid "Unlimited" plan let it train on everything you upload and generate. Protection only comes with an enterprise contract.
• Free tiers almost always train on you. Free Google AI Studio, Kling's free queue, and Runway's free plan all use your content. If your work matters, use a paid tier.
• Nothing here runs on your own computer. All five are cloud-only. For fully offline/private generation, you'd need an open-weight model (see the last section).
• Seedance and Kling are Chinese-owned, so your uploads are processed under Chinese data law. Fine for most hobby/portfolio work; weigh it carefully for anything sensitive or client-related.

Do you actually need an enterprise agreement?

Enterprise agreements sound serious and expensive — and for a two-person studio, they often are. The honest answer is that most small creators don't need one, as long as they pick the right platform and pay for it.

What to ask when you reach out

A short, specific email beats a sales call. Try asking:

• "On which plan, exactly, do you commit in writing not to use my uploads or generated videos to train your models?"
• "Is a no-training / data-processing agreement available below the full enterprise tier?"
• "What's the minimum commitment or monthly cost for that protection?"
• "Where is my data stored and processed, and can I choose the region?"
• "Do I get full commercial rights to what I generate, with no watermark?"
• "Can I delete my uploaded content and have it removed from your systems on request?"

Platform profiles

Google Veo — the safe default

For a small team, Veo is the lowest-drama choice. On a paid Gemini API or Vertex AI plan, Google contractually does not use your prompts, uploads, or outputs to train its models — and you don't need an enterprise deal to get that; it's the default on paid usage. You can also pick a region (including the EU) if where your data lives matters. The one catch: the free Google AI Studio tier does train on what you submit, so treat it as a sandbox for throwaway tests only.

Kling — budget-friendly and protected

Kling is a pleasant surprise for small budgets: its paid API agreement explicitly says it won't use your data to train or improve its models — a stronger written promise than Runway gives at the same price level, without an enterprise contract. The trade-off is jurisdiction: Kling is owned by Kuaishou, a Chinese company, so your uploads are processed under Chinese data law regardless of the contract. Fine for original animation that isn't sensitive; think twice for regulated client work. Avoid the free tier — generations can land in a public queue.

Runway — best looking, read the fine print

Runway is widely loved for cinematic quality and editing tools, and it's US-based. But the catch every small creator should know: Runway's Terms of Use state that your inputs and outputs may be used to train and improve its models — and that applies to all non-enterprise tiers, including the paid Unlimited plan. Paying for Runway doesn't, by itself, protect your animations. If you love the look and produce original IP, a negotiated no-training/enterprise agreement is the only way to turn training off. Otherwise, do protected work on Veo or Kling and use Runway only for content you don't mind contributing.

Seedance — capable, but opaque on data

Seedance (by ByteDance) produces impressive results and is reachable through BytePlus or services like fal.ai and Replicate. The problem for a careful creator is transparency: there's no clearly published, accessible no-training commitment for ordinary users, and like Kling it runs under Chinese data law. Treat it as a place to experiment with style — not the home for original characters you intend to own and monetize. If you want to use it seriously, ask ByteDance/BytePlus about a data-processing agreement first.

Gemini Omni — promising, but wait

Gemini Omni is Google's newest model, announced at I/O in May 2026 — a single model handling text, images, audio, and video together, with conversational editing. Right now it's live in Google's consumer apps, but the developer API isn't generally available yet. Practical advice: don't build your workflow on Omni yet. When its paid API launches, it's expected to inherit the same no-training protection as Veo — at which point it becomes a strong option. Until then, use Veo.

If you want full privacy: run your own

None of the five platforms can run on your own hardware — they're all cloud services, so your uploads always leave your machine. If your top priority is that your original work never touches someone else's servers, the only real answer is an open-weight model you download and run yourself.

• Open-weight video models like Wan, CogVideoX, or Mochi can be downloaded and run on a capable GPU, entirely offline.
• Quality generally trails the top commercial models, but the gap keeps narrowing — and nothing you make ever leaves your computer.
• This takes more technical setup and hardware, so it's best for teams comfortable with a bit of tinkering.

For most two-person studios, the sweet spot is a paid plan on a platform that contractually protects you (Veo or Kling). Keep the self-hosted route in mind for your most sensitive, signature work.

What about Figma Weave?

Figma Weave (formerly Weavy, acquired by Figma in October 2025) is different from the five platforms above. It isn't a single AI model — it's a node-based canvas that routes your work to many third-party models (Veo, Kling, Seedance, Runway, and more) in one place, with built-in editing tools. So the real question isn't "is Weave safe?" It's "is the model I picked inside Weave Verified?"

Verified vs. Unverified models

Figma labels every model in Weave as either Verified or Unverified, and that label tells you how your uploads are handled.

Here's the useful twist: a model that would train on you by default going direct can be protected when you run it through Weave on a Verified node, because Figma's contract overrides the default. Runway is the clearest example — risky direct, but safe as a Verified model in Weave. Always confirm the badge before uploading original characters, since verified status changes as Figma signs new contracts.