FFMPEG COMMAND

Extract audio from a video with FFmpeg

-vn drops the video; the codec you choose decides the rest. Copy the existing AAC when you only need the file, transcode to MP3 or WAV when a tool requires it.

Verified with FFmpeg 9.0.1 on 26 August 2026. Search intent: ffmpeg extract audio.

the command

ffmpeg -i input.mp4 -vn -c:a libmp3lame -q:a 2 output.mp3

Keeps only the audio and encodes it with LAME at VBR quality 2, roughly 190 kbps on typical content. The 12 s test file produced a 12.0 s MP3 with the original 48 kHz sample rate.

Flags, explained

-vn: Disable video: no video stream is written to the output.
-c:a libmp3lame: Encode audio with the LAME MP3 encoder.
-q:a 2: LAME variable bitrate quality, 0 (best) to 9. 2 is the usual choice for near-transparent MP3. Use -b:a 192k for a constant bitrate instead.
-c:a copy: Copy the audio stream as it is. The output container must accept the codec (.m4a for AAC).
-c:a pcm_s16le: Signed 16-bit little-endian PCM, the standard WAV format.
-ac 1 -ar 16000: One channel, 16 kHz sample rate.

Variants

Copy the AAC track without re-encoding

shell

ffmpeg -i input.mp4 -vn -c:a copy output.m4a

Most MP4 and MOV files already carry AAC audio. Copying it into an .m4a is instant and lossless; the .mp3 extension would not work here because the MP3 container cannot hold AAC.

Uncompressed WAV for editing

shell

ffmpeg -i input.mp4 -vn -c:a pcm_s16le output.wav

16-bit PCM at the source sample rate. Large (1.15 MB for 12 s of mono 48 kHz here) but no generation loss.

Mono 16 kHz WAV for speech-to-text

shell

ffmpeg -i input.mp4 -vn -ac 1 -ar 16000 -c:a pcm_s16le speech.wav

The format most transcription models expect. -ac 1 mixes down to mono and -ar 16000 resamples; the test file came out at 16,000 Hz, 1 channel, 384 KB.

Only a section of the soundtrack

shell

ffmpeg -ss 3 -t 5 -i input.mp4 -vn -c:a libmp3lame -q:a 2 segment.mp3

-ss and -t before the input select 5 s starting at 3 s. Output duration: 5.000 s.

Opus for the smallest files

shell

ffmpeg -i input.mp4 -vn -c:a libopus -b:a 96k output.opus

Opus at 96 kbps sounds as good as MP3 at much higher bitrates, but plays in fewer places (browsers and modern players, not older phones).

Pitfalls we hit

You cannot copy AAC into an .mp3 file: FFmpeg stops with "Invalid audio stream. Exactly one MP3 audio stream is required." Use .m4a with -c:a copy, or re-encode with libmp3lame.
Re-encoding AAC to MP3 loses a little quality and gains nothing if you only need the audio in a player. Copy when you can.
WAV output is large: 16-bit stereo at 48 kHz is about 11.5 MB per minute.
A video with several audio tracks yields only the first by default. Use -map 0:a:1 to pick the second.

How we verified it

Run with FFmpeg 9.0.1 on 26 August 2026 against a 12 s H.264 file with a mono 48 kHz AAC track. MP3 -q:a 2: mp3, 48,000 Hz, 12.000 s, 51,411 bytes. AAC copy: aac LC, 194,696 bytes. WAV: pcm_s16le, 1,152,078 bytes. Speech WAV: 16,000 Hz, 1 channel, 384,078 bytes. Segment: 5.000 s. Opus 96k: 178,108 bytes. AAC copied to .mp3 failed as described.

The same edit as a typed operation

extract_audio takes a format (mp3, wav or others) and returns the file: the codec, container and bitrate pairing is decided for you, so the AAC-into-MP3 mismatch cannot happen.

MCP tool: extract_audioPOST /api/v1/tools/extract-audio

request body

{
  "inputs": [{ "id": "main", "url": "https://example.com/video.mp4" }],
  "options": { "format": "mp3" }
}

Run extract audio in the browser Connect MCP API docs

FAQ

How do I extract audio from a video without losing quality?

Copy the existing stream with -c:a copy into a container that accepts it (.m4a for AAC, .mp3 for MP3). Any re-encode to a lossy format loses something; WAV keeps the decoded audio intact.

What bitrate should the MP3 have?

-q:a 2 (roughly 190 kbps VBR) is transparent for most listeners. Use -q:a 4 or -b:a 128k for speech, -q:a 0 or -b:a 320k when the source is high-quality music.

How do I get audio in the format Whisper or another transcription model wants?

Mono, 16 kHz, 16-bit PCM WAV: the speech variant above. Smaller files transcribe just as well.

Can I extract audio from a specific time range?

Yes: put -ss (start) and -t (duration) or -to (end) before -i, as in the segment variant.

Related commands

All FFmpeg commands