Downloads · 30 days
0
charlesw09/CLEAR-mask-free-video-subtitle-removal-CogvideoX
CLEAR-mask-free-video-subtitle-removal-CogvideoX is a text-to-video model from charlesw09. Use it when you need video from a text prompt. It is set up for diffusers. The card lists the license as apache-2.0.
This repository releases LoRA + expanded input-projection weights for video-to-video subtitle removal on top of zai-org/CogVideoX-2b.
Downloads · 30 days
0
Access
Public
Updated Apr 2, 2026
Repo size
119 MB
Likes
5
Public
Click a slice to open those files.
.pt119 MB · 100%
From the Hugging Face model README
This repository releases LoRA + expanded input-projection weights for video-to-video subtitle removal on top of zai-org/CogVideoX-2b.
Disclaimer: This is a supplementary experiment from the CLEAR project. The main paper results use Wan2.1-Control; this CogVideoX-2b variant is not expected to match that baseline. It is shared for reproducibility and comparison.
CogVideoX-2b is originally text-to-video. For conditioning, the first-stage conv input is expanded:
patch_embed.proj: Conv2d(16 → 1920, …)patch_embed.proj: Conv2d(32 → 1920, …)
Inference concatenates noisy latent and subtitle latent along the channel dimension before the transformer, consistent with training.
zai-org/CogVideoX-2b.cogvideox_2b_CLEAR_lora_checkpoint.pt locally.export MODEL_PATH="/path/to/CogVideoX-2b"
export CHECKPOINT="/path/to/cogvideox_2b_CLEAR_lora_checkpoint.pt"
bash scripts/inference_cogvideox_2b.sh \
--input_video /path/to/video_with_subtitles.mp4 \
--output_dir ./output