Openai Whisper

• You need to transcribe audio or video files locally without an API key or usage costs.
• You want privacy-preserving transcription where audio never leaves your machine.
• You need subtitle/caption files (.srt) generated directly from audio.
• You are processing a batch of recordings and want full control over model size vs. speed trade-offs.

• You need real-time or streaming transcription — Whisper CLI processes files, not live audio streams.
• You need a managed cloud transcription service with guaranteed uptime and SLAs.
• Your machine has very limited disk space or RAM and cannot accommodate the model weights (even the smallest models require several hundred MB).

Whisper runs entirely on your own hardware. No OpenAI account, no API key, and no data sent to external servers — just download the model weights once and transcribe offline.

Supports txt, srt, vtt, tsv, and json output formats via the --output_format flag, making it easy to generate captions, subtitles, or plain transcripts from a single command.

Pass --task translate to automatically translate non-English speech into English text, powered by the same model weights used for transcription.

Choose from a range of models (tiny, base, small, medium, large, turbo) to balance transcription speed against accuracy. The default on this install is turbo. Models are cached in ~/.cache/whisper after the first download.

Drop a recorded .mp3 or .m4a file into Whisper with --model medium --output_format txt to get a clean text transcript without uploading sensitive conversations to a third-party service.

Use --output_format srt to produce a subtitle file ready to import into video editors or media players, ideal for content creators captioning their recordings locally.

Run --task translate on foreign-language audio to get an English transcript in one step, useful for researchers or journalists working with multilingual source material.