XaiJu
Mickmumpitz
Mickmumpitz

patreon


[ADVANCED WORKFLOW] Shoot EPIC MOVIES with this FREE AI Tool!

Thank you so much for your support! This video took absolutely forever to make and it would not have been possible without you! 

You can download the advanced workflow, example files and reference videos below. 

The example outputs have the metadata in them so you can just drag and drop them into ComfyUI to recreate my prompts and workflow. 

You can find the INSTALLATION GUIDE here: https://docs.google.com/document/d/1UALpsT8EM4gbBZSIMcedsOskhAxEOb3IB6e7QUfEKWE/edit?usp=sharing

[ADVANCED WORKFLOW] Shoot EPIC MOVIES with this FREE AI Tool!

Comments

Hi there, How can I solve this error? "Cannot copy out of meta tensor; no data! Please use torch.nn.Module.to_empty() instead of torch.nn.Module.to() when moving module from meta to a different device." https://prnt.sc/5CPwuJujkBj1

Göktuğ Solmaz

For anyone that is reading this: He DOES answers, but he prefers discord. He literally fixed my problems in a few minutes when i asked him on discord... So hop on discord if you have issues or questions!

Selina Druifmans

Hello i have same laptop. You used the oosted files and did not get any compatibility issue? And what are the versions of cuda pytorch and confyui used? Thanks for sharing

Gartfortheworld

CogVideoSampler torch._scaled_mm is only supported on CUDA devices with compute capability >= 9.0 or 8.9, or ROCm MI300+ video frame count 137 i am using RTX3090 card

polaraju p

Is it possible to make img2vid same way ?

Badum

Thank you so much! I don't know what i am doing wrong but when i play the queue only the original video renders, i cant find the result anywhere, also it doesnt preview in the workflow, seems like the process stops before finishing... does someone know how can i fix it? I've tried to redownload and import again, but nothing happens. Thanks

Raquel López

마법사의 호모로 변신하거나 버그가 있는 워크플로에 대한 수정 사항을 게시해도 여전히 '수정'이 작동하지 않아 비용을 지불해야 할 때에도 움찔하지 않습니다. 이 아이는 언제 움찔할까요? 아마도 인공지능 게이 포르노에서 호빗들이 자신을 회색이 아닌 게이 간달프로 쳐다보는 것을 볼 때겠죠. 오, 사실 그는이 똥에 흥분합니다. 그는 자신의 똥 후원금으로 사람들을 사기 치는 병신입니다. btw - its the same text with "he does not cringe..." but only in korean

McLovin

just to remind you because i think the message got lost by all the other guys posting their error messages you never respond to - dont bother banning me, i already downloaded everything and canceld my membership :)

McLovin

dude that is so cringe turning yourself into gandalf like that, and really standing behind it like it so normal lolololol

McLovin

作者你好,我一直卡在#415 FILM VFI这个节点上 错误的提示是“PytorchStreamReader failed reading zip archive: failed finding central directory”,不管我如何下载film_net_fp32.pt这个文件,这让我非常困扰,是否能协助我解决这个问题,谢谢。

larry leng

Hello, I can´t install CogVideo on my Macbook. Does anyone know if it´s possible to do it?

Bladimir Salazar

can not run the other two video cause it do not detect the face?

Lee Mark

Hunyuan just made updates allowing it to run natively in ComfyUI. Has anyone tried to swap out Cog for Hunyuan?

Ramon Rodgers

[ONNXRuntimeError] : 3 : NO_SUCHFILE : Load model from E:\ComfyUI_windows_portable\ComfyUI\models\liveportrait\landmark.onnx failed:Load model E:\ComfyUI_windows_portable\ComfyUI\models\liveportrait\landmark.onnx failed. File doesn't exist

Jovanni Hogan

I haven't gone too deep with them personally but I was checking out experimenting with these CogX Loras https://github.com/Nojahhh/cogvideox-loras/blob/main/LORA_MODELS.md

Conrad Curtis

Thanks Mick, works fantastically for realistic clips, if I'm trying for an anime style what would be the best way to achieve this?

Pete Lumby

In the Down(load) CogVideo Model section I changed the quantization setting to fp8_e4m3fn

Pete Lumby

For those who don’t have an RTX 4090, this issue arises because certain operations require GPUs with compute capability 8.9+. Even GPUs like the RTX 3090 (compute capability 8.6) are not fully compatible. If you’re using an older GPU, you may encounter similar errors. To work around this, make sure you're using a compatible PyTorch version for your CUDA setup and look for alternative nodes or workflows that don’t rely on these unsupported operations

Ali fi

Hi Anh, advanced version work with 4090

Ali fi

Not yet, but there is a solution. However, I won’t be implementing it because I have multiple scripts on my PC that rely on my current configuration. Here's the solution for reference: I'm sure you will have difficulty changing your config especially getting xformers to work. **************************** Since you have an RTX 3090 (compute capability 8.6), the issue occurs because certain features like torch._scaled_mm require compute capability 8.9+, which is specific to GPUs like the RTX 40 series. Solutions for Your RTX 3090 Use a Compatible PyTorch Version Ensure you have PyTorch installed for CUDA 11.7: bash Copy code pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117 Disable Features Requiring Compute Capability 8.9/9.0 If a specific model, node, or library (like xformers) uses unsupported operations, try: Downgrading xformers to a stable version: bash Copy code pip uninstall xformers pip install xformers==0.0.20 --no-cache-dir Using alternative model nodes in ComfyUI that do not rely on unsupported CUDA features. Verify Model and Node Compatibility Check if the model or workflow you’re using explicitly requires GPUs with compute capability 8.9/9.0. If so, seek alternative models compatible with your RTX 3090. Force CPU as a Temporary Fix If GPU operations fail and you need a workaround, you can run ComfyUI in CPU mode: bash Copy code python main.py --cpu ************************************************************* The error you're encountering in ComfyUI indicates that a specific operation (torch._scaled_mm) being called by PyTorch is only supported on newer GPUs with compute capabilities >= 8.9 or 9.0. This issue typically arises when: You are using a CUDA version or PyTorch version incompatible with your GPU. The GPU's compute capability doesn't meet the requirement for the operation. A model or feature in ComfyUI requires newer GPU hardware. Steps to Fix the Issue 1. Check Your GPU Compute Capability Run this command to identify your GPU model and compute capability: bash Copy code nvidia-smi Cross-reference your GPU model with its compute capability using this list: CUDA GPUs and Compute Capabilities. Compute Capability 8.9/9.0: NVIDIA GPUs like RTX 40 Series (e.g., RTX 4090) or later. If your GPU's compute capability is lower (e.g., 7.5 for RTX 20 Series), the operation will fail. 2. Verify Your PyTorch and CUDA Versions Ensure your PyTorch version matches your CUDA installation. Run: bash Copy code python -c "import torch; print(torch.version.cuda)" python -c "import torch; print(torch.__version__)" If you're running CUDA 11.7 (your installed version), ensure you install PyTorch compatible with CUDA 11.7: Correct Installation: bash Copy code pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117 If you have CUDA 12.0+, install PyTorch for CUDA 12.0: bash Copy code pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118 3. Downgrade xformers or Torch Features If you're using xformers or custom nodes that require newer GPU features, it might cause issues: Uninstall and reinstall xformers with a compatible version: bash Copy code pip uninstall xformers pip install xformers==0.0.20 --no-cache-dir 4. Fallback to CPU (Temporary Fix) If you are unable to resolve the issue for now and want ComfyUI to run, you can force it to use the CPU: Modify your ComfyUI launch command: bash Copy code python main.py --cpu 5. Upgrade Your GPU (If Required) If your GPU compute capability is below 8.9/9.0, some newer operations in PyTorch or Stable Diffusion may not work. You would need to upgrade to a GPU like RTX 40 Series.

Ali fi

I am experiencing the same issue on my end. Did you find a resolution to this? I also have RTX 3090 24GB VRAM.

Jared Rowe

Try running update_comfyui.bat; fixed it for me. Note that seems to do more than update through the manager.

James

Hi, Could you please clarify the difference between the advanced version and the free one? I am using an RTX 3090 with 24GB VRAM, and while it successfully renders the free workflow, I encounter the following error when trying to use the advanced workflow: torch._scaled_mm is only supported on CUDA devices with compute capability >= 9.0 or 8.9, or ROCm MI300+ Could you provide details on what additional features or requirements are included in the advanced version that might be causing this issue? Thank you!

Ali fi

Your Pc : 4090 ?

Phuong Can

Thanq

Denys Los

I'll test 3060

Denys Los

Same, I am having issues as well

fitz Unit

awesome workflow! running well on a 3080ti 16gb laptop. question: In an older workflow I used for AnimateDiff a lot, I was able to use IP adapters to maintain consistent characters. Can this be done here?

Chris "Broadway" Romero

For some reason I can't get the CogVideo Decode to work.

thekriskooper

merci !!

laurent cantinieau

This is where it's at...thanks Mick!

John Nguyen


More Creators