File Transcription Skill

Transcribes audio or video files with Gemini through the project's AI proxy. No API key needed — billed to the project's credits per token, like a chat call (the last line prints the credits spent). Languages are auto-detected; mixed languages inside one recording are fine.

Default engine: gemini-flash-latest — keeps cross-talk and hard passages, follows a vocabulary list, labels speakers. MODEL=gemini-3.5-transcribe switches to Google's dedicated STT model: cheaper and gives word-level timestamps, but it silently skips hard passages of multi-speaker recordings, so use it for clean single-speaker audio only.

When to use

• Transcribe an audio file (MP3, WAV, OGG, FLAC, M4A, …)
• Transcribe a video file (MP4, MKV, MOV, AVI, WebM, …)
• Meeting / interview / lecture / podcast → text, optionally with who said what