Both renders came off one rtmpose pass, so the two files differ by the encoder and by nothing else. Both sit against the same 5 Mbit/s cap, which is why they are the same size: what the WebM codec buys here is picture, not compression. Look at the metrics panel and the bend badges — thin, sharp text over moving water is where a codec gives up first.
| file | codec | size | bitrate |
|---|---|---|---|
| dan.h264.mp4 | H.264 / MP4 | 16.7 MB | 5.15 Mbps |
| dan.vp9.webm | VP9 / WebM | 17.3 MB | 5.32 Mbps |
The transport drives both. Seeking is exact — every render carries a keyframe
every half second — and ◀ Frame steps one frame of 60.
compare-videos
H.264 on top, VP9 below, one file. Written by
compare-videos out/dan.h264.mp4 out/dan.vp9.webm, which is itself
a second encode over the two — so judge the codecs from the pair above, and
use this one for stepping through the stroke with both in view.
--layout blendThe two laid over each other at 50/50. Where the encoders agree the frame looks normal; where they diverge the picture beats against itself.