Published by PitchChanger · Updated · How these guides are made
How to use this workflow
Inspect the file outside this site
Use a local media inspector that lists streams. Note audio stream 1 and any later audio streams; this site has no stream picker.
Match the first stream to your task
Confirm its language or musical cue before extracting. If the wanted track is second, create a local single-stream copy in another editor first.
Extract and check the MP3
Open the saved MP3 and listen near the beginning and later cue. A successful download proves one stream was processed, not that all tracks were kept.
The first audio stream is the only stream selected here
The extractor maps the first audio stream in the chosen local video and encodes it as a 320 kbps MP3. It does not offer language selection, multi-track export, WAV output, or a YouTube URL importer.
For a hypothetical file with stream 1 = director commentary and stream 2 = rehearsal mix, this tool will produce commentary. The extension and the player’s default audio choice do not override that order.
External inspection command: ffprobe -v error -select_streams a -show_entries stream=index,codec_name:stream_tags=title,language -of json "source.mp4". Audio ordinal zero is the first audio stream, even if its overall stream index is 1 because video occupies index 0. FFmpeg’s explicit -map 0:a:0 selects that first audio stream, rather than whichever stream a player defaults to.
Swipe or scroll sideways to read all columns.
| Source layout | Result here | Useful next step |
|---|---|---|
| One intended audio stream | That stream is encoded as MP3 | Extract and listen |
| Commentary first, mix second | Commentary is encoded | Make a local single-stream copy externally |
| No audio stream | Nothing can be extracted | Obtain a source with audio |
An actual two-stream check: first is not the default
We created a two-second black video with two original synthesized tones: first audio stream 440 Hz, second audio stream 880 Hz. The second was marked as the default. These are test signals, not music or spoken language.
On 22 October 2026, the real browser Audio Extractor saved an MP3 from that video. Decoding the result to 48 kHz mono and estimating frequency from interpolated positive-going zero crossings in the middle second gave 439.997 Hz. That identifies the first 440 Hz stream, despite the second being marked default. This checks one controlled file, not every container or browser.
Swipe or scroll sideways to read all columns.
| Audio ordinal | Overall stream index | Tone | Default flag |
|---|---|---|---|
| 0 (first) | 1 | 440 Hz | No |
| 1 (second) | 2 | 880 Hz | Yes |
Use two recognizable cues to check the result
Before extraction, write down a spoken line, count-in, or instrument entrance near the start and one later cue. After download, locate both in the MP3. This checks content identity as well as whether the file opens.
Do not infer stream order from a filename such as “final.mp4”. A video container can include several audio streams, and another player may choose a different default.
Keep the workflow within its limits
Choose one nonempty local video up to 250 MB. Browser codec support, decoded memory, and storage can still reject a file below that size. Processing stays in the browser, while the site and engine files still need to load.
Use a capable external editor when you must select another stream, preserve multiple tracks, extract lossless audio, or work from a remote URL.
Common questions
Can I choose a second language track?
No. This extractor has no audio-stream selector and processes only the first audio stream.
Does an MP4 always have one soundtrack?
No. It can have none, one, or several audio streams. Inspect it if the exact track matters.