extract-audio

Exact captured content · path skills/extract-audio/SKILL.md · sha256 f7de70b0ac83a24f… · revision 9f12c4154ebdde0f99ad7bf24222130dde1110d6

Download the exact captured file

---
name: extract-audio
description: Get the sound out of a video or recording - save it as MP3, M4A, WAV or FLAC, cut a part, make it evenly loud, or prepare a 16 kHz mono WAV for transcription - and pull embedded subtitles out as SRT. Use when the user asks to extract audio, rip the soundtrack, turn a video into a podcast or MP3, get audio for transcription, or save the subtitles of a video.
---

# Audio (and subtitles) from video

Keep the original file. Name results after it (`meeting.mp4` → `meeting.mp3`) and do not overwrite anything.

## 1. Look first

```sh
ffprobe -v error -show_entries stream=index,codec_type,codec_name:stream_tags=language,title -of compact "$IN"
```

This lists the audio and subtitle tracks. If there are several audio tracks (languages, commentary), ask which one; select it with `-map 0:a:N`.

## 2. With ffmpeg

| Goal                                   | Command                                                                                     |
| -------------------------------------- | ------------------------------------------------------------------------------------------- |
| MP3, good quality                      | `ffmpeg -i in.mp4 -vn -c:a libmp3lame -q:a 2 out.mp3`                                       |
| M4A without re-encoding (AAC source)   | `ffmpeg -i in.mp4 -vn -c:a copy out.m4a` (if it fails, the track is not AAC: encode to MP3) |
| WAV for editing                        | `ffmpeg -i in.mp4 -vn -c:a pcm_s16le out.wav`                                               |
| FLAC (lossless, smaller than WAV)      | `ffmpeg -i in.mkv -vn -c:a flac out.flac`                                                   |
| Only a part (from 1:30, 2 minutes)     | `ffmpeg -ss 00:01:30 -t 120 -i in.mp4 -vn -c:a libmp3lame -q:a 2 part.mp3`                  |
| Evenly loud (podcasts, talks)          | `ffmpeg -i in.mp4 -vn -af loudnorm=I=-16:TP=-1.5:LRA=11 -c:a libmp3lame -q:a 2 out.mp3`     |
| For transcription (Whisper and others) | `ffmpeg -i in.mp4 -vn -ac 1 -ar 16000 -c:a pcm_s16le speech.wav`                            |
| Second audio track                     | add `-map 0:a:1` before the output name                                                     |

## 3. Subtitles

| Goal                        | Command                                               |
| --------------------------- | ----------------------------------------------------- |
| First subtitle track to SRT | `ffmpeg -i in.mkv -map 0:s:0 out.srt`                 |
| A given language            | `ffmpeg -i in.mkv -map 0:s:m:language:eng out.en.srt` |
| SRT ↔ VTT                   | `ffmpeg -i in.srt out.vtt`                            |

Only text subtitles (SubRip, ASS, mov_text, WebVTT) convert to SRT. Picture-based ones (PGS on Blu-ray, DVD subtitles) need OCR; say so instead of producing an empty file. Videos without subtitle tracks have nothing to extract: offer transcription of `speech.wav` instead.

## 4. Without ffmpeg

With the Converter Studio MCP tools, `convert` the video with `to: "mp3"`, `"m4a"`, `"wav"` or `"flac"`; subtitle files convert with `to: "srt"` or `"vtt"`. Taking only the sound costs 1 credit per started hour of the recording (a video converted to another video costs about 1 credit per 3 minutes at 1080p). Without either, https://converter.studio does it in the browser.

## Report

Give the output's length, format and size, and which track was used. For subtitles, say how many lines came out and the language.