search

nateraw / stable-diffusion-videos

Create 🔥 videos with Stable Diffusion by exploring the latent space and morphing between text prompts
Apache License 2.0
4.4k stars 421 forks source link

Update Replicate interface to include latest changes here #63

Open billybailey47 opened 1 year ago

billybailey47 commented 1 year ago

Any way to upsample/scale the images before the video is created like with Google Colab? Sucks to pay for something which outputs inferior quality.

nateraw commented 1 year ago

Not sure exactly what you mean - can you be a little more specific about what you're trying to do?

billybailey47 commented 1 year ago

Hi, thanks for the comment!

I'm using the Replicate UI (https://replicate.com/nateraw/stable-diffusion-videos) to produce videos with stable diffusion. There doesn't seem to be an option in the UI to check a box to upsample like there is with Google Colab. Will this feature be added as there is a massive quality difference between the two platforms?

E.g Replicate outputs grainy video, collab does NOT when you click to upsample.

I attach two screenshots to visually show this!

Screenshot 2022-10-06 at 09 23 33 Screenshot 2022-10-06 at 09 22 19
billybailey47 commented 1 year ago

@nateraw If there is a way, please guide me! I'm probably missing something.

nateraw commented 1 year ago

While I do technically "own" the replicate demo, I haven't been able to update it myself because I would need to have a GPU machine to build the docker image and push up the changes. That demo was made with a very early version of the code (see #9), and doesn't include a lot of the latest updates here.

I don't really have a clear path forward to solving this issue due to my lack of local GPU, so not sure what else I can do (maybe @chenxwh can help? ❤️)

chenxwh commented 1 year ago

Hi @nateraw, if you update the Cog predict.py to the latest, I can push the demo for you :D