comfyanonymous
Please support for ChronoEdit
Feature Idea
https://huggingface.co/nvidia/ChronoEdit-14B-Diffusers ChronoEdit-14B enables physics-aware image editing and action-conditioned world simulation through temporal reasoning. It distills priors from a 14B-parameter pretrained video generative model and separates inference into (i) a video reasoning stage for latent trajectory denoising, and (ii) an in-context editing stage for pruning trajectory tokens. ChronoEdit-14B was developed by NVIDIA as part of the ChronoEdit family of multimodal foundation models. This model is ready for commercial use.
Existing Solutions
No response
Other
No response
Feature Idea
https://huggingface.co/nvidia/ChronoEdit-14B-Diffusers ChronoEdit-14B enables physics-aware image editing and action-conditioned world simulation through temporal reasoning. It distills priors from a 14B-parameter pretrained video generative model and separates inference into (i) a video reasoning stage for latent trajectory denoising, and (ii) an in-context editing stage for pruning trajectory tokens. ChronoEdit-14B was developed by NVIDIA as part of the ChronoEdit family of multimodal foundation models. This model is ready for commercial use.
Existing Solutions
No response
Other
No response
it's already working. I already tested it in Comfy. The only thing he needs is to somehow make the length 2 frames, since the standard requires at least 5, but for this model you only need 2.
Model: https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/chrono_edit_14B_fp16.safetensors
Workflow:
Model: https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/chrono_edit_14B_fp16.safetensors
Hi, there is a ScaleRope node, can't find it via custom manager, is it unpublished custom node?
It is only available in git comfyui, it will be in stable next week.
Model: https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/chrono_edit_14B_fp16.safetensors
Workflow:
Is there fp8 model?
Model: https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/blob/main/split_files/diffusion_models/chrono_edit_14B_fp16.safetensors Workflow:
Is there fp8 model?
https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/tree/main/ChronoEdit
Is it possible to visualize the editing trajectory? If I display all 5 images generated, apart from the first one which is the input original, the other 4 are the resulting image.
Is it possible to visualize the editing trajectory? If I display all 5 images generated, apart from the first one which is the input original, the other 4 are the resulting image.
I use the 'video combine node' save and to see how it chooses the last image. Sometimes, the images it skips are much better than the final result, especially if you use a WAN T2V LoRA with it. Some LoRAs affect the image in steps that is skipped in final output. I extract the frame I want and upscale it.
@rzgarespo i don't think we are talking about the same thing. On the Nvidia presentation page of ChronoEdit they show that it's possible to recover the "thinking" process of the model to see how it came to the conclusion. Scroll down to the: "Temporal Reasoning Visualization" section. This is a video, which's first frame is the input and the last frame is the output.
@rzgarespo i don't think we are talking about the same thing. On the Nvidia presentation page of ChronoEdit they show that it's possible to recover the "thinking" process of the model to see how it came to the conclusion. Scroll down to the: "Temporal Reasoning Visualization" section. This is a video, which's first frame is the input and the last frame is the output.
Thank you for the link. That is exactly what the video combine node achieves. it makes a video clip out of images rendered for each step. So, at the end, there is one image as the final result, and an MP4 to see the thinking process. It's quite interesting, especially when a cat is involved like. prompt: "a cat sitting on the couch. " , Set the steps to 65 :)
@rzgarespo then please share the workflow how you managed to render that video. As I said I only get a video with multiple repeated frames of the final image if I save the result of VAE Decode with Video Combine (from VHS nodes).