AI Content Policy

Ozu uses artificial intelligence to generate visual and audiovisual production assets from your creative inputs. This AI Content Policy describes how AI is used within the Service, your rights and responsibilities regarding AI-generated content, and the current legal landscape around AI outputs.

1. How AI Is Used in Ozu

Ozu integrates multiple third-party AI systems to power its pipeline. The mix of models changes as we add and retire them. The current list, what we send each provider, and each provider’s training posture, lives in Privacy Policy §9. That table is the source of truth — this page does not keep a second copy that can rot.

In plain terms, today that stack includes:

Direct

• Anthropic Claude — script analysis and the Ozu assistant (chat)
• xAI Grok — images, video, speech-to-text, and text-to-speech
• Google Gemini / Veo / Lyria / TTS — images, video, music, and speech
• OpenAI — images and 360° panoramas
• ElevenLabs — TTS, voice cloning, sound effects, and music (replica-sensitive)
• Replicate — music analysis and dialogue/foley separation (Demucs)
• Blockade Labs — skybox / environment generation
• World Labs — 3D environments from panoramas (not currently reachable in the app)

Via the EvoLink generation gateway

• Kling — video
• Seedance / Seedream — video and images
• Seed Audio — voiceover / TTS (replica-sensitive, same as other voice tools)
• Midjourney — images
• Suno — music
• Qwen — TTS (replica-sensitive)
• VideoRetalk — lip-sync

Ozu, your project-aware creative assistant, reviews assets, drafts prompts, and maintains project context. It is powered by Anthropic Claude.

2. Content Ownership

We assign to you all rights we hold in Output Content. We do not claim ownership of your generated characters, storyboards, environments, video, voice, music, or other assets.

That is not a promise the output is copyrightable. Under current US copyright law, as interpreted by the US Copyright Office (Copyright and Artificial Intelligence Part 2, January 2025), works generated solely by AI without sufficient human creative input are generally not eligible for copyright protection. Human-authored elements — your script, your selection and arrangement, substantial modification, and your edits — may still be protectable. We do not promise that any film or asset is registerable with the Copyright Office. Consult an intellectual property attorney for your specific use case.

Commercial use of outputs is also limited by the applicable model vendor’s terms. Midjourney in particular has its own commercial-use tiers. World Labs’ free-tier terms may provide that users do not own output. You’re responsible for the terms of the providers that produced a given output. Ozu cannot claw back a training license a vendor takes under its own terms.

See Terms of Service §6.2 for the full disclosure.

3. Your Creative Direction Matters

Ozu is designed to maximize your creative control over AI outputs. You are not pressing a button and receiving random images. You are:

• Writing the script that defines every character, location, and scene
• Answering Ozu's consultation questions to fill in visual gaps
• Reviewing and approving every generated asset at the Director Review Board
• Composing shots in 3D space with Camera Control
• Selecting, reordering, and timing frames in the Timeline

This level of creative direction may strengthen a claim in the human-authored elements of the resulting work. It does not, by itself, make a purely generated frame copyrightable. Ozu is a tool in your creative process, not a replacement for it.

4. AI Content Disclosure

We recommend that you disclose AI involvement when publishing or distributing content generated through Ozu, particularly in contexts where disclosure is legally required or expected by the audience.

Emerging regulations in several jurisdictions require disclosure of AI-generated or AI-assisted content, especially in commercial contexts. While requirements vary by jurisdiction, transparency about AI involvement is good practice and may become a legal obligation in your market.

5. Content Moderation

AI generation requests are subject to the content policies of our upstream providers as well as our own Acceptable Use Policy. Requests that violate these policies may be blocked or result in account action.

Ozu does not manually review every generated image. We rely on upstream provider safety filters combined with our own monitoring for policy compliance.

A provider safety-system decline of a single generation is not the same thing as us terminating your account for an AUP violation. Credits for a declined generation are returned; see the Refund Policy and Terms §3.2.

6. No Training on Your Data (by Cloud Creator)

Ozu does not use your scripts, generated images, video, audio, project data, or any user content to train AI models. Your content is processed to deliver the Service and is not contributed to any training datasets of our own.

Some optional features route your content to providers whose terms permit training on what is submitted. We identify those features before you use them. Using them is your choice. The current per-provider marks are in Privacy Policy §9.

7. Limitations of AI Generation

AI-generated content is probabilistic. Results may vary between generations, even with identical inputs. Generated characters may not be perfectly consistent across multiple frames. Generated environments are artistic interpretations, not photographic reproductions. Ozu provides tools to manage consistency (character sheets, approved references, assistant QA) but cannot guarantee pixel-perfect reproducibility.

7.1 Resemblance to real people

Generated voices, faces, and performances may incidentally resemble real people. That resemblance is not intended as a portrayal, endorsement, or digital replica of any particular person. We do not warrant that outputs are unlike any real person.

Naming an actor or film as shorthand for a type (“a Brad Pitt for this role”) is allowed. Asking the Service to generate that person, or uploading their photo, voice, or footage as reference, is not. If a named reference is converted into attributes, those attributes have to describe a type, not a recipe for one identifiable person.

Voice generation — Seed Audio (via our generation gateway), ElevenLabs, and Qwen TTS — is replica-sensitive. Recorded reference video (including video-to-video) of anyone other than the account holder needs their e-signed likeness-consent form, stored with the project. A checkbox from the account holder is not enough. A signed form is not a union clearance.

A disclaimer of unintentional resemblance is not a substitute for rights clearance and is not a defense under publicity or replica laws.

8. BYOK (Bring Your Own Key)

Studio+ users who connect their own API keys run generation on those keys instead of Ozu credits. Ozu’s servers still make the HTTP call. We are an intermediary for that request; it does not go from your device directly to the provider, and your key is not exposed in the browser.

The provider’s content policies, billing, training, and retention terms for your account apply to that traffic. If you want zero-retention or similar settings, you configure them on your provider account. We don’t log the content of BYOK calls.

Full terms: BYOK Addendum.

9. Professional and union productions

If you are working under a guild or union agreement, that agreement may restrict use of generative-AI literary material, synthetic performers, and AI footage. Those rules are between you and the guild. They are not part of these Terms, and we don’t police them for you.

10. Changes to This Policy

The legal landscape around AI-generated content is evolving rapidly. We will update this policy as new laws, regulations, and legal precedents emerge. Material changes will be communicated to registered users via email.

11. Contact

Cloud Creator LLC
c/o Northwest Registered Agent, 30 N Gould St, Ste N, Sheridan, WY 82801
[email protected]

Revised August 13, 2026: updated the provider stack (including Seed Audio via gateway) and pointed training details at Privacy §9; matched ownership language to Terms §6.2 and USCO Part 2 (January 2025); corrected BYOK (Ozu servers make the call); added vendor commercial-use, resemblance/no-endorsement (not a liability shield), actor-as-type vs replica, vid2vid e-sign, and optional union/GAI notes; standardized the registered-agent address to Ste N.