Open
Conversation
FeepingCreature
force-pushed
the
feature/anima
branch
from
7d60e0d to
ce6acad
Compare
- resources.py: Add Arch.anima enum member with text_encoders (qwen_3_06b), supports_attention_guidance, search paths for TE/VAE, required resource IDs - workflow.py: Load separate CLIP (omnigen2 type) and VAE for anima in load_checkpoint_with_lora() - comfy_client.py: Add qwen_3_06b to text encoder model discovery Co-authored-by: Forge AI (github.com/FeepingCreature/forge) <noreply@forge-ai.invalid>
FeepingCreature
force-pushed
the
feature/anima
branch
from
ce6acad to
0679bd4
Compare
Acly
reviewed
Apr 20, 2026
| return Arch.qwen_l | ||
| if string == "qwen-image": | ||
| return Arch.qwen | ||
| if string == "anima" or (string == "unknown" and "anima" in filename): |
Owner
There was a problem hiding this comment.
I think this should already be detected, but probably as lumina2, can you check?
Since base Lumina 2 isn't supported, we might as well map it to Anima.
| match (model_info.format, model_info.quantization): | ||
| case (FileFormat.checkpoint, Quantization.none): | ||
| model, clip, vae = w.load_checkpoint(model_info.filename) | ||
| if arch is Arch.anima: |
Owner
There was a problem hiding this comment.
This shouldn't be needed if the model is correctly located in diffusion_models / unet folder.
Technically people could make a Anime "checkpoint" with integrated TE/VAE and this would prevent it from being loaded... not sure how important that is, but since other models don't do this atm I'd remove it for now.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Anima is a promising up and comer! Not super slow like Flux, but also with much better prompt understanding than anything OpenCLIP based, it's still in testing but it looks to be the first model I'd cautiously say is a straight improvement over SDXL.
ComfyUI supports it ootb, so there's almost nothing to do. There are only two finicky details: there's no model class for it yet so we have to recognize by filename, and the preview releases don't bake in VAE and TE (because they're frozen) so they have to be loaded separately even tho it's a standard safetensors load node.
Also of course there's no controlnets yet, though a first derived model already popped up, so it's probably just a matter of time.
(thanks claude for code!)