AI Disclosure
Last updated
Heady makes videos with AI. This page explains which parts are generated, which models we use, how we mark the videos you export, and what you are responsible for when you publish them.
1. What in a Heady video is AI-generated
| Part | Generated by AI? | How |
|---|---|---|
| Outline and script | Yes, then approved and edited by you | Large language model drafts; you can edit any line and approve the script |
| Images | Yes | Image model generates the first frame; every later frame is an AI edit of it |
| Motion | No model-generated video | Depth parallax, camera moves and transitions are computed from the images |
| Motion mode graphics | Yes, as code | A language model writes the animated scene (charts, titles, timelines, counters); our renderer draws it |
| Narration | Yes | Text-to-speech model reads your approved script |
| Subtitles | Text from your script; timing by AI | Speech recognition times each word against the narration |
| Chapter cards, lower-thirds, progress bar | No | Fixed templates filled with your chapter titles and text |
| Music | [PLACEHOLDER: state the music source and whether any of it is AI-generated] | |
| Chapter timestamps, SRT, description | Timestamps and SRT from the video; description drafted by AI |
2. Models we use
At the date of this page, Heady uses these models. AI requests go through OpenRouter to the provider that runs each model, except where we say a model runs on our own servers.
| Task | Model | Provider |
|---|---|---|
| Outlines, scripts, descriptions, Motion mode scenes | OpenAI GPT-6 Luna family [PLACEHOLDER: confirm models and fallback models] | OpenAI, through OpenRouter |
| Images | Gemini 2.5 Flash Image | Google, through OpenRouter |
| Standard narration | Gemini 3.8 Flash Lite TTS | Google, through OpenRouter |
| Narration on [PLACEHOLDER: plans] | Kokoro-82M (open source) | Our own servers |
| Premium narration | ElevenLabs [PLACEHOLDER: model name] | ElevenLabs, through OpenRouter [PLACEHOLDER: confirm route] |
| Word timing for subtitles | faster-whisper (Whisper speech recognition, open source) | Our own servers |
We will update this table when the models change. Our Privacy Policy explains what data these providers receive.
3. How exported videos are marked
Under Article 50(2) of the EU AI Act, which applies from 2 August 2026, providers of AI systems that generate synthetic audio, images or video must ensure the output is marked in a machine-readable format and detectable as artificially generated or manipulated.
Every video you export from Heady carries:
- [PLACEHOLDER: e.g. a C2PA Content Credentials manifest, signed by Heady, embedded in the MP4, stating that the video was generated with AI and naming Heady as the tool];
- [PLACEHOLDER: e.g. an imperceptible watermark in the video or audio that survives common re-encoding].
[PLACEHOLDER: state whether a visible label is added, on which plans, and where in the frame.]
[PLACEHOLDER: state whether the description Heady exports includes a line saying the video was made with AI tools.]
You can check a Heady video's marks with [PLACEHOLDER: verification tool, e.g. a public Content Credentials verifier].
These marks are part of how we meet the law, and part of how platforms recognise AI content. Removing or stripping them is prohibited by our Acceptable Use Policy and Terms of Service. Note that some platforms and editing tools remove metadata when a video is re-encoded; the watermark is designed to [PLACEHOLDER: describe robustness honestly].
4. What you need to disclose
Machine-readable marks do not replace the disclosures you are responsible for as the person who publishes the video:
- Deep fakes. Article 50(4) of the EU AI Act requires anyone who publishes AI-generated or manipulated images, audio or video that resemble real people, objects, places or events and would falsely appear authentic to disclose that the content is artificially generated or manipulated. For evidently artistic, creative, satirical or fictional work, the disclosure can be made in a way that does not spoil the work.
- YouTube. YouTube asks creators to disclose realistic altered or synthetic content, such as a realistic scene that did not happen or a real person appearing to say something they did not. It does not require disclosure for clearly unrealistic or animated content, or for using AI to help with an outline or script. Check YouTube's current rules when you upload.
- Other platforms have their own labels. Follow the rules of each one.
Heady's illustrated style will often fall outside "realistic", but that depends on your video. When a video could be taken for real footage of a real event or person, disclose it.
5. Human review
Heady keeps a person in the loop. You approve the script before production, can edit every line of it first, and review the finished video before you publish it. AI can still make mistakes, including factual errors. You are responsible for checking a video before you publish it; see our Terms of Service.
6. Contact
Questions about this page or about AI marking: [PLACEHOLDER: contact email].