Lossless compression stores a file more efficiently without discarding anything: decompressing returns the original bytes, or in video terms the original pixels, exactly. FFV1, lossless H.264, and formats like PNG for still images work this way.
The catch is the ratio. Lossy video compression routinely achieves 10:1 or better. Lossless video compression achieves something in the region of 2:1 on favourable content and much less on noisy footage. That is useful for an archive and useless for an upload limit — a 300 MB clip is not reaching 20 MB losslessly, and no tool can make it.
Where lossless genuinely applies:
- Archival masters you intend to re-encode from later, where each future encode should start from the original rather than from a lossy copy.
- Intermediate files in editing, where footage may be decoded and re-encoded several times and generation loss would accumulate.
- Remuxing, which is lossless in the sense that matters most often: the encoded streams are copied unchanged into a different container. Fast, free, and the right answer when the goal is compatibility rather than size.
One place the word is used loosely: "visually lossless" is not lossless. It means lossy compression at a quality level where the loss is meant to be invisible — typically a low CRF. Our own measurements put libx264 at CRF 18 between 96 and 97 VMAF, which is excellent, and is not 100.