Nvidia Cosmos In ComfyUI - AI Diffusion Model For Video Generation - Setup Tutorial

[@BenjisAIPlayground]

All model files, text encoder, vae, Folder location , etc... listed in here : www.patreon.com/posts/120128927?

[@CHATHK]

You need to mention vram requirements first

[@content1]

Thank you Benji, I am missing the EmptyCosmoLatentVideo node? Any ideas. (I am running the latest Comfyui ;) Thank u!

[@FusionDeveloper]

Some other video workflows have the typo of 25 fps, so always check and change it to 24.

[@BenjisAIPlayground]

I think its still new, so Comfy keep try and error for this model.

[@damarcta]

thanks for the quick update 👍👍🖐🖐🖐🖐

[@BenjisAIPlayground]

Glad it help :)

[@gigend]

how much vram minimum?

[@gjohgj]

Superb video!

[@FusionDeveloper]

Thanks for the update

[@RhysAndSuns]

nice digging, thanks

[@Carl-md8pc]

Thanks Benji

[@BenjisAIPlayground]

glad it help :)

[@kalakala4803]

It have potential but need some fine tuning.

[@BenjisAIPlayground]

yes, like Jensen Huang said, it's using Nemo to feed in data. So I guess we can based on this and add our own dataset on top of it.

[@pdjay8912]

감사합니다:)

[@ashertique4651]

Thank you for this. Another informative video. I've updated ComfyUI and am searching for the CosmosImageToVideoLatent node and nothing is coming up. Can't find it. Can you point me to a link to install this node please?

[@ashertique4651]

Never mind. I didn't realize the "update" in Manager is not as up to date as running the Update.bat file. Node loaded.

[@BenjisAIPlayground]

yes, please refresh the web browser if needed. I find out that sometimes the Web UI are caching some old Web Ui stuff as well.

[@ashertique4651]

@@BenjisAIPlayground Good to know, thanks.

[@mikea8858]

Comfyui master update just dropped fir this

[@JamBassMusic]

Nice video, Benji! NVIDIA produces amazing products, but the issue is they’re a bit slow with support for other platforms like Replicate or ComfyUI...

[@BenjisAIPlayground]

First few generation test, I was kind of disappointed of human motions. Because NVIDIA claim this series of models are using real world physic. But it doesn't looks like it. Not sure, keep trying some settings.

[@YoungBloodyBlud]

@@BenjisAIPlayground please let us know if its faster than hunyuan or ltx video and how is the t2v