When running on a folder of mp3s, nearly all of them report loudness range larger than the loudness range target about ffmpeg-normalize HOT 1 CLOSED

In principle, you can ignore the warning and let the algorithm perform the normalization with dynamic mode. If the results sound okay to you, then it's fine. This will perform “compression” on the input signal to reach the target loudness, thus avoiding large volume jumps.

For other cases, i.e. to preserve the dynamic range and just shift the loudness, users want linear mode. For this, the settings have to be chosen such that the target loudness range is equal to or larger than the input loudness range. That's why there's a warning, because the filter chooses its mode implicitly — and it's not always clear when this happens.

If you want linear mode (which I think you don't, according to your use case), to increase the target loudness range to the input loudness range (i.e., to avoid normalization with dynamic mode), you can pass the --keep-lra-above-loudness-range-target option. This will always automatically set the target loudness range to the input's loudness range to avoid reverting to dynamic mode.

PS: Over time, options have become more complex to handle special cases that are not directly covered by the filter. If you have a suggestion on how to make these warnings or the README more user-friendly, please let me know (or propose a PR).

from ffmpeg-normalize.