Refactor video generation and model loading processes to utilize from_pretrained methods for VideoEncoder and VideoDecoder. Update denoising functions to include a cfg_rescale parameter for improved artifact reduction. Ensure consistent dtype handling across audio and video processing, enhancing precision and aligning with PyTorch behavior.

2026-01-23 17:39:02 +01:00
parent 02bfa228d9
commit df753312c7
3 changed files with 119 additions and 151 deletions
--- a/mlx_video/models/ltx/video_vae/decoder.py
+++ b/mlx_video/models/ltx/video_vae/decoder.py
@@ -560,3 +560,7 @@ class LTX2VideoDecoder(nn.Module):
            chunked_conv=use_chunked_conv,
            on_frames_ready=on_frames_ready,
        )
+
+
+# Backward-compatible alias
+VideoDecoder = LTX2VideoDecoder