Your media and model training.
Your media and model training. We do not use your photographs, films, words, voice or music to train or improve any model.
This commitment is ours. It does not describe what the companies on our subprocessors list do.
What the companies we use may do. Google, OpenAI and Anthropic read your photographs and plan your film. Their terms for our paid accounts say they do not train models on what we send. The account holder confirmed on 2026-10-12 that no training permission was given to Google, that data sharing with OpenAI is off, and that the Anthropic permissive opt-in is off.
fal turns your photographs into moving scenes, with its own models and those of other companies. fal's terms let it use anonymised data derived from what we send to develop its AI models. Kling's published terms let it train on what it receives. xAI's say it does not, without permission. The account holder confirmed on 2026-10-12 that no permission to train on our material was given to fal or to xAI. For MiniMax, and for Google's model behind fal, we found no clause either way.
ElevenLabs speaks the narration, composes music and clones your own voice, where we switch those on. It receives the music prompt on PROD: the film's own music direction, as text. On TEST it receives the narration text, a voice clone's recordings and the music prompt. Its terms let it train on what it receives. The account holder confirmed on 2026-10-12 that the account is the same free-tier account on both stages, and that it has not opted out of "Data use", so it may train on what we send it.
Each of these companies keeps what we send for a time its terms set. Our subprocessors list names those we use directly on our live service.