Retake feature in ComfyUI
Hi all,
I've just seen at the end of the official web (https://ltx.io/model/ltx-2) the Retake feature.
Is it possible to use this feature with the Open Sourced model? In that case, any ideas how to use it inside ComfyUI? Or anyone can share a template?
Thank you. Bests,
Miguel
You could try using canny/depth LoRA for that.
I'm trying with it now. I thought there was a special way to fo the retake feature. Let's see if I can get it working.
Thank you.
According to the Retake tutorial at https://youtu.be/bzU82Olv8_M?si=W2R2JL3LJ5wTeA6y
It seems to blanks the frames that need to be changed 🤔 i guess it's just frames interpolation similar to VACE or FLF.
Yes, in the API section of the documentation: https://docs.ltx.video/api-documentation/api-reference/video-generation/retake
It says that you set the input video, the start frame and duration to replace that part of the video, but it doesn't look a "start/end frame" workflow, as the movement of the camera and the image of the video is respected. Maybe internally it does a depth or canny LoRa plus start/end frame workflow.
May be it tooks more than 1 frames at the first/last marked frame to get the motion 🤔
There is this Temporal Inpainting workflow (which i believe to be this Retake thing) at https://comfyui.nomadoor.net/en/basic-workflows/ltx-2/