Guys, nothing phenomenal. I’m presenting you with a convenient workflow for replacing a character’s actions in a video. With a well‑written prompt, the characters do the same things they usually do in video, only with some changes that you want. And sometimes it looks absolutely amazing! The total generation time is up to 1 minute. Let's take a fresh look at our favorite videos))

Rules:

1. Mentally divide the desired video according to your compute power into 1–4 fragments of n seconds each (5-15sec).

2. Prepare stop‑frames at these seconds, where the result you want will occur. You can use QWEN or Z‑Image.

3. Place the stop‑frames in the control frame section, starting with the first original frame, and upload the video to the video section. Switch the right number of Generation videos.

4. Describe the n‑number of prompts and what happens in them, taking your changes into it.

5. Select a checkpoint and Turbo lora for generation — be sure to use ref2va.

6. Select LORA so that the generation is suitable. Be careful, as LORA often ruins the entire generation.

7. RUN