Why does the render take a few minutes?

AutoYop transcribes your audio, calls Claude AI to plan every scene, fetches stock footage from multiple sources, renders animated graphics with a headless browser, and assembles everything with ffmpeg. Each step takes real time — but the result is a fully produced video.

What API keys do I need?

At minimum: Anthropic (Claude) for visual planning. For voiceover generation: ElevenLabs. For stock footage: Pexels and/or Pixabay. For image search: Serper. Optional: fal.ai for AI-generated images when stock runs out. Go to Settings →

How long can my script be?

AutoYop handles scripts up to ~12 minutes of spoken audio. Longer scripts are batched automatically so Claude can plan each section without hitting token limits.

The render got stuck — what do I do?

Click Cancel and try again. The most common cause is a slow network fetch for stock footage or an animation render that timed out. The system automatically retries and falls back to footage if an animation fails. If it keeps failing, check View logs for the specific error.

Why are my API keys gone after an update?

Make sure your API keys are also set as environment variables in Railway (ANTHROPIC_API_KEY, ELEVENLABS_API_KEY, etc.). The app re-bootstraps them from env vars on every deploy so they're never lost.