The slow tool, and why
Every other video tool here is a stream copy: packets move from one container to another without ever being decoded. That is why trimming and muting finish in seconds regardless of file size.
Converting cannot work that way. Changing the codec means decoding every frame back to pixels and compressing them again with a different algorithm. Video encoding is one of the most computationally demanding things a consumer device does, and running it as WebAssembly in a browser tab costs roughly ten times what the native binary would.
The honest summary: this works, it produces a correct file, and it is slow. Before starting one, it is worth checking whether the Trimmer or the Muter already does what you need.
Containers and codecs
These are two different things, and confusing them is the source of most format frustration. A codec — H.264, VP8, AV1 — is the compression algorithm applied to the picture. A container — MP4, WebM, MKV — is the file format that wraps the compressed streams together with timing information.
The same H.264 video can sit inside an MP4 or an MKV. A player that refuses your file may be rejecting the container, the codec, or both, which is why "convert to MP4" sometimes fixes a problem and sometimes does not.
- MP4 — H.264 video, AAC audio. The universal default.
- WebM — VP8 video, Vorbis audio. Open, royalty-free, slower to encode.
- MKV — H.264 in Matroska. Extremely flexible, poorly supported on the web.
How CRF works
Constant Rate Factor asks the encoder for consistent perceived quality rather than a fixed data rate. A static shot of a wall gets very few bits because it needs very few; a fast pan through foliage gets many more. The result is that quality stays even across the video instead of collapsing whenever the scene gets complicated — which is exactly what a fixed bitrate does.
The scale runs roughly 0 to 51, where lower is better. Around 20 is visually indistinguishable from the source for most material. Around 26 is a sensible default. Past about 32 the softening becomes obvious on anything detailed.
Generation loss
Every re-encode discards information permanently. Converting an already-compressed video means the new encoder is working from footage that has already had detail removed — and it cannot tell the difference between real detail and the artifacts of the previous encode, so it faithfully preserves both.
Convert once, from the best source you have. Chaining conversions compounds the damage in a way that is very visible after two or three passes.