Settings

Whisper Transcription

Larger models are more accurate but slower. "medium" is a good default for GPU.
float16 for GPU, int8 for CPU.

LLM Summary (Optional)

Leave blank for provider default. API keys come from .env (OPENAI_API_KEY, ANTHROPIC_API_KEY, GEMINI_API_KEY, or legacy LLM_API_KEY). Ollama needs no API key. Reprocess picker in the gallery only offers providers whose keys are set.
Only used when provider is Ollama. Default works for Docker Desktop (Mac/Windows). On Linux use http://172.17.0.1:11434 or your host IP.
Adds a "Transcript (Corrected)" section to every markdown file. Disable if LLM cost is a concern.

Submission Pipeline

When off, videos submitted via the extension are only downloaded. Use the gallery to select which ones to transcribe. Useful when your library is large and whisper is the bottleneck.

Auto-cleanup

After a saved-list video finishes transcribing, move it to the list below. The next extension poll handles the move (non-destructive).
The Instagram saved list the extension will move transcribed videos into. Created automatically if it does not exist.

Transcript Storage

Local saves markdown files to the mounted data volume. Google Drive requires a connected Google account.
Only used when storage is set to Google Drive.