A true peak is the maximum of the reconstructed continuous waveform, measured in dBTP with 4x oversampling per ITU-R BS.1770 — real program material routinely hides 0.5 to 1.5 dB of inter-sample peak above the loudest stored sample, so a plain sample-peak limiter set to the ceiling still ships overs. This tool runs a closed loop: the input true peak is measured with FFmpeg’s ebur128 (peak=true, the BS.1770 Annex 2 oversampled method); if it is already under the ceiling the file is simply converted losslessly; otherwise FFmpeg’s lookahead alimiter is applied with its threshold set BELOW the ceiling by an inter-sample safety margin (the limiter only sees samples, so the margin reserves room for the peaks between them). The output is then re-measured with the same oversampled meter, and while any true peak remains above the ceiling the margin grows in 0.5 dB steps and the pass repeats — the loop only accepts a verified reading, never the limiter’s own claim. The limiter always runs with auto-level off (otherwise it renormalises the output back up and defeats the ceiling) and lookahead delay compensation on. Output is written as lossless PCM (32-bit float WAV or 24-bit FLAC) so the tool itself adds no codec overshoot; measure and limit after any sample-rate conversion, per AES TD1008. Ceiling presets follow current delivery practice: -1.0 dBTP for streaming and EBU R128 broadcast, -2.0 dBTP where heavy lossy encoding follows (Amazon Music, Netflix, low-bitrate HE-AAC), -1.5 dBTP conservative streaming, -0.3 dBTP for lossless/hi-res releases. Integrated loudness is reported before and after so the loudness cost of the limiting stays visible.