Add multi-prompt video generation support from prompt file.

[glmcdona]

Adds arguments to hv_generate_video.py to allow for specifying a file with a list of prompts to generate videos from:

• --prompts_file: Path to text file with prompts (one per line)
• --increment_seed: Increment seed for each processed prompt
• --repeat_count: Repeat generations of each prompt X times

Also a couple minor documentation tweaks to make it more clear for users.

[GarbageHaus]

I tested this out and it's really excellent. When generating small videos like 288x480 it seems to increase the generation speed by around 50% as a rough estimate.

If I could make a humble request to have the unique identifier that is inserted in the front of the filename to be appended to the end instead? This makes organization easier since the first part contains a date.

[glmcdona]

I tested this out and it's really excellent. When generating small videos like 288x480 it seems to increase the generation speed by around 50% as a rough estimate.

If I could make a humble request to have the unique identifier that is inserted in the front of the filename to be appended to the end instead? This makes organization easier since the first part contains a date.

Good idea, thanks for the feedback! I think this will be more consistent with the original output naming convention. +1 to the speed of generation, this greatly speeds up generation of videos, since it doesn't require reloading the models and re-merging Loras for each video generation! Seems to about double the speed for my case and was the reason I wanted to implement this.

[kohya-ss]

Thank you for this pull request!

However, I would like these functions to be compatible with gen_img.py in sd-scripts repo. Is that possible?

https://github.com/kohya-ss/sd-scripts/blob/sd3/gen_img.py

If that is difficult, I would like to implement these features myself, referring to this PR.