What happens
A GIF export runs about 28× slower than an MP4 export of the same project on the same machine.
Measured on 2.0.0-rc.1 (Windows 11, AMD Radeon iGPU, 125 % scaling), on a 56.9 s recording with a zoom, a trim, a speed region and a full-camera region:
| Export |
Frames |
Wall time |
Throughput |
| MP4 720p 30 fps |
1495 |
~18 s |
~84 fps |
| GIF Small (480p → 864×480) 15 fps, loop |
748 |
~4 min |
~3 fps |
The progress readout advanced at a steady ~9 frames every 3 s, with an ETA of about 5 min at the start.
The output itself is correct. ffmpeg decodes it without errors, and ffmpeg and Pillow give byte-identical frames on six sampled frames: 748 frames, 49.87 s, NETSCAPE2.0 loop.
Ask
Profile where the GIF path spends its time: frame readback, palette quantisation, or LZW encode. Nothing in the GIF path changed between rc.1 and main, so this is not a regression of the release candidate. It makes GIF export impractical beyond short clips.
What happens
A GIF export runs about 28× slower than an MP4 export of the same project on the same machine.
Measured on 2.0.0-rc.1 (Windows 11, AMD Radeon iGPU, 125 % scaling), on a 56.9 s recording with a zoom, a trim, a speed region and a full-camera region:
The progress readout advanced at a steady ~9 frames every 3 s, with an ETA of about 5 min at the start.
The output itself is correct. ffmpeg decodes it without errors, and ffmpeg and Pillow give byte-identical frames on six sampled frames: 748 frames, 49.87 s,
NETSCAPE2.0loop.Ask
Profile where the GIF path spends its time: frame readback, palette quantisation, or LZW encode. Nothing in the GIF path changed between rc.1 and
main, so this is not a regression of the release candidate. It makes GIF export impractical beyond short clips.