Skip to content
Memo is in private beta, by invitation. Join the waitlist

AI usage policy

Last updated: September 29, 2026

We use state-of-the-art AI models to analyze your content. Memo never trains AI models on your inputs or outputs. Image and video generation providers handle your content under the conditions in our Terms of Service.

How we analyze private content

We use OpenRouter to route requests to third-party AI models while ensuring security and data isolation. Data shared with OpenRouter is used only during processing and discarded immediately afterward without logging, retention, or sharing.

Data is never shared with the original model provider for training purposes. We ensure providers adhere to rigorous data isolation and security standards to prevent unauthorized access. Read OpenRouter's privacy policy here.

How we analyze public content

Public content means public URLs (including media) that can be accessed by anyone with a browser and a public link.

We may use additional third-party AIs and proxies to analyze public content. These services may log requests for security and diagnostic reasons. We do not associate any requests made to our AI services with individual user accounts or personal identifiers.

Models we use today

  • Image analysis: Gemini 2.5 Flash via OpenRouter — for auto title, description, and deep analysis (palette, prompt, composition, lighting, mood).
  • Image generation: Nano Banana, GPT Image, Seedream, Muse Image, Recraft and Krea via fal.ai. Soul 2 via Higgsfield.
  • Video generation: Kling, MiniMax, Wan, and Veo via fal.ai. Seedance and Genjutsu via Higgsfield; any Seedance job Higgsfield cannot take runs on fal.ai.
  • Studio chat: Claude (Anthropic) via the Anthropic API.