Cogvideox#13402

Merged

comfyanonymous merged 15 commits intomasterfrom

Apr 29, 2026

Commits on Apr 23, 2026

Initial CogVideoX and SparkVSR support

kijai
authored and
Talmaj Marinc
committed
Remove breaking code, logging etc.
Talmaj Marinc
committed
Remove sparkvsr related code.
Talmaj Marinc
committed
Utilize use_learned_positional_embeddings in forward pass of CogVideoX.
Talmaj Marinc
committed
Fix mutable input parameter.
Talmaj Marinc
committed
Fix CogVideoX concat_cond to handle temporal dimension and normalize channel count
Talmaj Marinc
committed
Add CogVideoX 1.5 geometry defaults to I2V path
Talmaj Marinc
committed
Fixup ruff.
Talmaj Marinc
committed
Remove vae_backup.py
Talmaj Marinc
committed
Move cogvideo text encoder into a dedicated module.
Talmaj Marinc
committed
Cap encode chunks at the configured frame batch size.
Talmaj Marinc
committed
Cap encode chunks fix.
Talmaj Marinc
committed
Avoid pre-interpolating z for the full clip at every high-res stage.
Talmaj Marinc
committed
Fix cogvideox dtypes and ops.
Talmaj Marinc
committed
Revert dtype to float32 to increase quality of video output.
Talmaj Marinc
committed