Image optional (t2v)

[neph1]

Maybe not so useful to many, but it might be interesting to some. Image is made optional, so you get a text to video workflow.

Also incorporates PR #158 so you can change resolution.

All output will be square.

[junweifeng]

This is very useful. The visuals are fantastic when used with multiple LORAs. The character likeness of Hunyuan LORA is superior to Wan2.1. That's why I've always used Hunyuan instead of Wan2.1 for text-to-video generation. Now with FramePack, the image quality is practically comparable to the original video.

[morganavr]

t2v is very useful for quickly trying different ideas.

[mocat15]

t2v is great for B-roll!

So the approach could be a drop-down approach to switch between i2v and t2v, or have a bypass switch.

[ronbere]

sorry for my ignorance? should I install your version, or just git your branch?

[neph1]

Either should work. The main branch has all of my PR's, but you can check this feature out here.

[rmcc3]

I keep getting this error.

  File "E:\FramePack\demo_gradio.py", line 20, in <module>
    from diffusers_helper.load_lora import load_lora
ModuleNotFoundError: No module named 'diffusers_helper.load_lora'```

[neph1]

Sorry. Left over from other branch