image2video

[DAigory]

is it possible to generate image to video?

[deepbeepmeep]

Not yet. We are still waiting for Tencent to release this feature.

[peteromallet]

@deepbeepmeep, I believe Skyreel based their code on diffusers and their approach is probably similar to what Tencent will do: https://github.com/SkyworkAI/SkyReels-V1

[peteromallet]

Thank you for all that you do @deepbeepmeep!

[mirix]

There is a very interesting model coming:

https://github.com/Saiyan-World/goku

And this has just been released:

https://huggingface.co/Wan-AI/Wan2.1-I2V-14B-720P

SkyReels does not even offer 720p.

[kakaxixx]

Looking forward to the adaptation to the Skyreels i2v model

[kwas-123]

How do I use this repo with comfyui?

[gituser123456789000]

How do I use this repo with comfyui?

Wondering the same.. but there's no mention of ComfyUI that I see (at least not to date)

In ComfyUI, you can use MutiGPU nodes and .gguf models to offload the models from VRAM to RAM, which frees up VRAM to push better quality models, higher resolution, longer duration, etc. Not sure how this HunyuanVideoGP competes with that. The numbers stated seem like this HunyuanVideoGP can perform better, but I'd need to see it actually used and results compared in what models were used, resolution, duration, etc.