Extract Audio from Video

Lossy formats (MP3, AAC, Opus) re-encode the audio, which means some detail is discarded. For most material the difference is inaudible at 192 kbps. For perfect quality choose WAV - it copies the PCM waveform exactly without re-encoding loss. Note that the source audio in a video is usually already compressed (AAC inside MP4), so the absolute ceiling is set by the source.

YouTube and most video sources store audio as AAC at roughly 128-192 kbps. Re-encoding from AAC to MP3 introduces a second lossy compression pass on top, called transcoding loss. To minimize the loss use the same bitrate or higher (192 kbps MP3 from 128 kbps AAC is fine), or pick WAV to skip the second compression entirely.

In theory yes (FFmpeg with `-c:a copy` writes the audio stream as-is), but the output container has to match the codec. AAC inside MP4 can be copied as M4A; MP3 inside MKV can be copied as MP3. We default to re-encoding because it produces a predictable, universally playable file. If you need a stream copy ask for it as a feature request.

AAC is the newer, more efficient codec. At the same bitrate AAC sounds noticeably cleaner than MP3, especially on cymbals, applause and other high-frequency material. The downside is compatibility - very old devices and some legacy players only handle MP3. For 2026 use cases AAC is usually the better technical choice; MP3 wins only on universal compatibility.

At 192 kbps a 60-minute MP3 is around 80-90 MB; at 320 kbps it is 130-140 MB. WAV at 44.1 kHz stereo is ~10 MB per minute, so a one-hour file is ~600 MB. Opus at 128 kbps is the smallest of the lossy options at ~55 MB per hour.

Most likely because the source video had higher-bitrate audio than the output target (e.g. a music video with 320 kbps source extracted to 192 kbps MP3 - the MP3 is smaller). Or you picked WAV which stores raw uncompressed samples. To make a smaller file pick MP3 or Opus at a lower bitrate.

Files travel over HTTPS to api.zeroutil.com (EU server). FFmpeg processes them locally on the server, returns a signed download URL, and both the input video and the extracted audio are auto-deleted after 15 minutes. We do not log content, do not transcribe audio, and do not retain anything beyond the deletion window.