Versions
int8: recommended. Fast, accurate, compatible with almost any GPU.
mxfp8: added for comparison. In theory (and according to nVidia PR) should be more accurate than int8, but in practice I was not able to spot any definitive advantages. A bit slower than int8, but still faster than original bf16. Compatible only with RTX 50xx series (Blackwell).
Performance on my setup
original bf16 (baseline): 2.20 it/s +0%
int8: 3.23 it/s +46%
int8 + torch compile (comfy core): 3.59 it/s +63%
int8 + turbo lora, cfg=1: 6.50 it/s +295%
int8 + turbo lora, cfg=1 + torch compile (comfy core): 7.55 it/s +343%
mxfp8: 2.58 it/s +17%
This is high quality int8 quantized version of base Anima v1.0 model. It retains ~90% of original model quality, but uses about 50% less VRAM and also runs faster on almost any nVidia GPU (AMD not tested). Nice trade-off, especially for low-end GPUs.
Can be used as a drop-in replacement for original Anima model in latest ComfyUI, no custom nodes required. If you have troubles running the model make sure that you updated both ComfyUI itself and its dependencies (e.g.pip install -U -r requirements.txt on manual linux install).
Converted to int8 / mxfp8 using convert_to_quant script.
Description
int8, ConvRot with dynamic group size, rowwise, SVD
Potentially better quality than ConvRot with fixed group size, but reported to be slower by some users.
FAQ
Comments (20)
In the ComfyUI Model Loaders Node
Does Int8 have to appear in the list of Weight_dtypes?
Because the model seems to load, and work fine.
But i have no clue if it's actually loading in Int8 quantitation or not.
If someone knows the awenser to this i would much appreciate it.
Thx.
@somedoby Did You tested with: https://github.com/Starnodes2024/comfyui-starnodes-modelconverter
This allows conversion directly from ComfyUi.
I was wondering how does MXFP8 vs int8 compare quality / speed / VRAM size.
That could be nice test opportunity.
Personally I wonder how Krea2 compares and maybe LTXv23
mxfp8 is a strange one. It should be better quality, but on my very surface level Anima testing I didn't noticed improvements compared to int8. Maybe it is because ComfyUI doesn't support ConvRot with mxfp8 (I think it should be possible, maybe just not implemented yet). Also mxfp8 was a bit slower. And I think mxfp8 is supported on RTX 5xxx only. So right now int8 looks to be a safer choice.
Thank you for uploading the model. By the way, what is the difference between v1 and v2?
when i testing v1 and v2 with forge neo, v1 is faster than v2.
Thanks for the report. I added warning to the model description, until I have time to test it myself.
I can confirm that v2 is way slower than v1, even slower than original Anima - don't know why.
There is any built in artist styles?
Tons of. https://animadex.net/?mode=artists
@somedoby How do you trigger it? With "by"or "@"?
@Hugh_Mungus "@"
@Twisowk I get beating trying to use this CK,i downloaded both tensors and so on,just generate some "statics" in ForgeUI NEO (Sorry,not a big fan of ComfyUI...)
@Hugh_Mungus , do you have text encoder and qwen-image vae? That model works differently than SDXL-based checkpoints. Visit official Anima checkpoint page for more information.(Suggested Resources)
@Twisowk Yep,i downloaded both files and the checkpoint,and i see a pattern:Diffusion in Low Bits when i select "int8"generate just noise and the other options generate the error "KeyError: 'int8_tensorwise'"
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.