Mix 5.1 or 7.1 sound down to stereo with ffmpeg on Windows: why the same -ac 2 comes out 7.7 dB…

Guide ID: art_01M7BKAPM66CBFEPQEX1SESAXC

Editorial publication checks are separate from community and evidence review. The evidence labels retain this revision’s reported inspection, execution and review status.

Guide title

Mix 5.1 or 7.1 sound down to stereo with ffmpeg on Windows: why the same -ac 2 comes out 7.7 dB louder as float than as 24-bit, and a script that sets the level by measurement

Coverage

Aimed at people holding surround sound (5.1 or 7.1) who want a stereo copy with a predictable level and no clipping. A PowerShell script picks the sound stream, applies ffmpeg's default down-mix matrix in floating point (front 1.0, centre and surround 0.707, LFE left out unless asked), measures peak and loudness, scales to a target peak, writes WAV, FLAC, M4A or a video with the picture copied, and checks channels, length, peak and picture packets. It ran on generated 5.1, 5.1(side) and 7.1 tone files with ffmpeg 9.0.2, including fifteen refusal cases; real film mixes and listening were not tested.

Example

The same ffmpeg -ac 2 gave a peak of -24.1 dBFS written as 24-bit integer and -16.4 dBFS written as 32-bit float (five equal channels), and on a 5.1 test clip -9.0 dBFS as 24-bit, -1.3 as float and +2.4 as AAC. The script set the peak of the stereo result to -1.0 dBFS (-3.0 for AAC, which overshot by 2.2 dB at -1.0).

Useful starting points

• The level of ffmpeg's default 5.1-to-stereo down-mix depends on the output sample format: integer output is normalised, float and AAC output is not.
• Measure the peak of the decoded result, and use a lower target peak when the result is AAC.
• The LFE channel is not mixed into the stereo result by default; add it deliberately if you want the bass.
• Listen to speech in the centre channel: it is mixed in at 0.707 like the surrounds.