ATTACH V2
A new function in the node pack. Built to raise the influence of Qwen 3.5 4B. It needs a little more memory than the old method. Direct development of Delta Qwen 3.5 4B. Source: Anima 2B — Qwen 3.5 4B Text Encoder.
This model has two modes. Both are selected in the new Clip loader.
3.8 1.1 base — the old mode, unchanged.
Qwen 3.5 strong — new. Slightly better frame reading, better scene layout. Test it.
Important: strong is still a test mode. It has its own problems, and a number of them are already known. If you hit one yourself, or have a fix in mind, write it in the comments.
Delta Mix 3.8B Qwen 3.5 - v5
Same line. v5 is the Attach V2 runtime. The bake is still DiT only. The Semantic Connector is still a sidecar, not inside the file.
The working path is the earlier one: 3.8 1.1 base, full frame. Same sidecar, same official connector path, same picture. If you only want what the previous version already did, stop here.
qwen 3.5 strong is the other CLIP format. It makes Qwen 3.5 4B the leading CLIP. Native 0.6B stays in the graph. This format does not use the 2560 sidecar as the conductor.
Samplers: dpmpp_2m / euler_a / res_multistep Schedulers: res_uniform / normal / beta
Warning
Attach V2 is not stable when you switch CLIP format on one live load. The problem is known. Part of it cannot be fixed in this node without breaking the path.
If the frame smears, drifts, or ignores the format you just set: change the loaded model, then load this checkpoint again, then set the format you want. Do not keep switching base / strong on the same load.
Delta Mix 3.8B Qwen 3.5 - v4
A recut of the same line, not a new family. v3 tightens frame stability and edge clarity: lines and object borders stay even instead of swimming. Same dual brief — human and furry.
Samplers: dpmpp_2m / euler_a / res_multistep
Schedulers: res_uniform / normal / beta
Delta Mix 3.8B Qwen 3.5 - v3
Anatomy and hands: fewer collapsed fingers and joint errors on the same test prompts.
Q35 path: the base holds pose / crop more steadily when the 4B encoder is in the graph. Stem (W): replaced. W is now a custom assembled 2B stack, not the previous floor.
Delta Mix 3.8B Qwen 3.5 - v2
Anima 3.8B mix for the official v1.1 chassis and Qwen 3.5 Connector v2.
Source: CircleStone Labs / lylogummy Anima 3.8B v1.1.
This bake fills W + Gen-1 + Gen-2 (Delta Mix overlay).
Connector v2 and the Q35 watch slots stay as in v1.1.
Qwen 3.5 is attached from a sidecar, not baked into this file.
Cleaner image than v1. Fewer W / Gen-2 clashes.
Fine detail and extra fingers still slip — for v3.
Delta Mix 3.8B Qwen 3.5 -v1
Anima 3.8B mix for the official v1.1 chassis and Qwen 3.5 Connector v2.
Source: CircleStone Labs / lylogummy Anima 3.8B v1.1.
The DiT is a Delta Mix overlay (stem + light Gen-1). Gen-2 insert slots and the Connector stay as in v1.1. Qwen 3.5 is attached from a sidecar, not baked into this file.
Downloads
This checkpoint
Sidecar anima_connector_only → same folder (or !Anima3.8/)
Nodes: ComfyUI-Anima-Delta-Q35-V2
Official runtime: comfyui-anima-3-8B (MIT) — Qwen 3.5 4B loader and Connector class
Text encoders from the v1.1 guide: Qwen 0.6B + Qwen 3.5 4B
Qwen 3.5 4B loader: GumGum10/comfyui-qwen35-anima. CLIP Loader V2 calls it for qwen 3.5 strong. Weights page: https://civarchive.com/models/2455272
Lora loader: ComfyUI-Anima-Delta-Q35-LoRA
UNET → DeltaMix Q35 LoRA → Attach V2 → Prompt V2Install
This bake is DiT only (~7 GB). The Connector and Qwen 3.5 4B are not inside the file.
This checkpoint → ComfyUI/models/diffusion_models/ (or models/unet/)
anima_connector_only.safetensors → same folder
qwen35_4b.safetensors → ComfyUI/models/text_encoders/ (or models/clip/)
qwen_3_06b_base.safetensors → same folder. CLIPLoader, type stable_diffusion
ComfyUI-Anima-Delta-Q35-V2 → ComfyUI/custom_nodes/
Both GumGum10 runtimes, same folder:
Also needed: native CLIP qwen_3_06b_base, same place as the 0.6B line.
2. Nodes (Git)
ComfyUI/custom_nodes/comfyui-anima-3-8B
https://github.com/GumGum10/comfyui-anima-3-8B
ComfyUI/custom_nodes/ComfyUI-Anima-Delta-Q35-V2
(zip on this page) or git clone https://github.com/bestluner-create/ComfyUI-Anima-Delta-Q35v2.git
ComfyUI/custom_nodes/ComfyUI-Anima-Delta-Q35-LoRA
https://github.com/bestluner-create/ComfyUI-Anima-Delta-Q35-LoRARestart ComfyUI after the three packs.
comfyui-anima-3-8B is the official runtime (4B loader + Connector class). Without it the Q35 graph will miss nodes.
Graph
UNETLoader this bake (~7.3G)
→ DeltaMix Q35 Attach V2 sidecar on 3.8 1.1 base
CLIP Loader V2 clip_format
→ DeltaMix Q35 Prompt V2 CLIP 0.6B + CLIP 4B
expanded → CFG / SamplerLoRA (optional):
UNET → DeltaMix Q35 LoRA → Attach V2 → Prompt V2Prompt Dual Delta: pos = quality tags + scene, neg = negative tags only.
Use expanded for sampling. native is 0.6B-only.
Extract once: DeltaMix Q35 Extract Connector on anima38B_11.safetensors. Do not extract again unless v1.1 changes.
This file does not include the Connector (~1.2G) or the 4B encoder.
Do not load *Q35bundle in this graph. Bundles are for the official v1.1 loader.
Sampler
Same stack as the v1.1 graph: RandomNoise → CFGGuider → SamplerCustomAdvanced + BasicScheduler. A plain KSampler is fine if the numbers match.
Steps 40. CFG 6. Denoise 1.0.
Sampler: er_sde, Euler, or whatever you already run on 3.8.
Scheduler: normal for this mix. beta if you want the stock v1.1 cadence.
Empty latent 1024×1024. 1920×1080 is stable.
Connector strength 1.0. Both Prompt nodes use the expanded output.
Do not match seeds between the official V2Loader graph and this one. The noise path is different. Compare seeds only inside the same graph.
Credits
V-5
Anima - CircleStone Labs
3.8B v1.1 chassis and Semantic Connector v2 - lylogummy
comfyui-anima-3-8B - GumGum10 (MIT)
Extract / Attach / Prompt - Delta Mix
Method: Anima Delta Mix — 2B into 2.9B / 3.8B
Custom - G1
Custom - G2
Custom - core
V-4
Anima - CircleStone Labs
3.8B v1.1 chassis and Semantic Connector v2 - lylogummy
comfyui-anima-3-8B - GumGum10 (MIT)
Extract / Attach / Prompt - Delta Mix
Method: Anima Delta Mix — 2B into 2.9B / 3.8B
Custom - G1
Custom - G2
DeMix - core
V-3
Anima - CircleStone Labs
3.8B v1.1 chassis and Semantic Connector v2 - lylogummy
comfyui-anima-3-8B - GumGum10 (MIT)
Extract / Attach / Prompt - Delta Mix
Method: Anima Delta Mix — 2B into 2.9B / 3.8B
Aura_Rising - G1
aniMurr - G2
Custom - core
V-2
Anima - CircleStone Labs
3.8B v1.1 chassis and Semantic Connector v2 - lylogummy
comfyui-anima-3-8B - GumGum10 (MIT)
Extract / Attach / Prompt - Delta Mix
Method: Anima Delta Mix — 2B into 2.9B / 3.8B
Aura_Rising - G1
aniMurr - G2
anima_aesthetic - core
V-1
Anima - CircleStone Labs
3.8B v1.1 chassis and Semantic Connector v2 - lylogummy
comfyui-anima-3-8B - GumGum10 (MIT)
Extract / Attach / Prompt - Delta Mix
Method: Anima Delta Mix — 2B into 2.9B / 3.8B
Aura_Rising - G1
Chosen-mix_anima - core
Description
FAQ
Comments (18)
Can I use this with Forge Neo? I'm using version 2.28, but I can't get it to stop producing black images.
Forge Neo 2.28 can run official Anima 3.8B + Qwen 3.5, but this file is not that bundle.
This checkpoint is a ComfyUI Q35 bake: UNET + Semantic Connector sidecar + native Qwen 0.6B together with Qwen 3.5 4B. Black frames almost always mean Forge loaded it as a plain 2B / SDXL-style path — wrong latent, no connector, or the 4B encoder in the wrong slot.
If you still want to try Forge:
Keep qwen_image_vae and qwen_3_06b_base in the top VAE / Text Encoder dropdown. Do not put qwen35_4b there.
Install the 3.8B extension: GumGum10/forge-anima-3.8B (needs Neo 2.28+). The 4B encoder is selected inside that extension, not in the top bar.
Shift around 3–4.5. Empty / 4-channel latent will stay black.
I build and test this page in ComfyUI portable, not Forge. If official Anima 3.8B v1.1 already goes black in your 2.28 install, this merge will too — fix the official graph first. ComfyUI is the supported path for this checkpoint.
@bestluner180 thanks
@bestluner180
369
I managed to solve it; actually, I just had to update Forge Neo to version 2.29 and reinstall the extension. I have to say the jump in quality is huge—I went from 2b to this, and honestly, I doubt I'll ever go back to Anima 2b. The model is truly incredible. Anyway, thanks a lot; your comments led me to try things that helped me find the solution.
@nutakugame007654 Good to hear 2.29 + a clean reinstall of the 3.8B extension fixed it. Glad the jump from 2B landed.
Model preliminary trial feedback:
Strengths:
● Improved ability to understand prompts, with reduced cases of missing objects and errors;
● The image output relies more on prompts, and the yield of good images has increased significantly.
Weaknesses:
● When the prompt exceeds 400 or 450 tokens, the generated image becomes blurry;
● Multi-character scenes look stiff, character movements tend to be similar, compositions and motions are formulaic with a noticeable AI look. The prompt handling has indeed been enhanced. It would be even better if the aesthetics could match those of anima aesthetics v11. Thanks for the creators' work!
Thanks for the trial notes — this matches what the Q35 stack actually does.
The stronger prompt follow and fewer missing objects come from the native Qwen 0.6B + Qwen 3.5 4B pair. That is the point of this branch. The blur after ~400–450 tokens is a limit of this text-encoder / CLIP path, not a hole in the bake. Past that length the tail of the prompt smears. Shorter prompts hold the frame better than longer ones.
Multi-character stiffness is real. Aesthetic v1.1 was used in earlier versions (V2/V3). On Qwen 3.5 a school shift toward Aesthetic v1.1 breaks the image, so V4 only keeps donors/schools that do not carry Aesthetic understanding. Official Aesthetic v1.1 is a different product; I am not treating it as a target to match on this page. I have not solved that compatibility yet.
Detailer stacks will not fix the token blur or the stiff multi-character look on this line. Clean plate + a style LoRA is safer than extra detailers.
Appreciate the write-up.
@bestluner180 Yes, your model ensures prompt adherence and stable composition, which are the most essential foundational capabilities. Other effects can be gradually improved on this basis or enhanced via LoRA, so it is not a major issue. I tried out 3.8b anima for the first time today and now have an intuitive understanding of it. This is indeed a groundbreaking achievement. Thank you again for your work!
Add some test details:
The first sampling produces highly stable images with great handling of image details and a high yield rate. The second sampling is very effective and natural in improving fine details. The upscale_models used is 4x_fatal_Anime_500000_G.pth, which delivers good results but slightly darkens the image.
If the prompt length exceeds 300 tokens, there is a 6% to 8% chance that the generated image will become blurry. It is recommended to keep it under 280 tokens. If secondary sampling is used, keep it within 240 to 260 tokens.
Hope this helps everyone!
Can I train lora for anima 3.8B and what programm i should use for it?
@KeMiliUs Sorry, I don't know much about this. Theoretically, you can use https://github.com/sorryhyun/anima_lora for training. However, to be safe, you'd better confirm it with the author.
@KeMiliUs
Yes, you can train a LoRA for the Anima family. I do not ship a 3.8B / Q35 training recipe myself.
Practical options:
sorryhyun/anima_lora — current dedicated trainer. Documented bases are official Anima (28 blocks) and Anima-2.9B (40). Depth is read from the checkpoint header. There is no published 52-block / 3.8B-Q35 guide in that repo. Treat 3.8 + Qwen 3.5 as untested there, not “just point at my file”.
Other community tools people use on official 2B: kohya anima_train_network, ComfyUI Anima LoRA Trainer, AnimaForge. Same caveat: they were built around official 28-block + Qwen 0.6B, not this Q35 bake.
Safer path on this checkpoint:
Train the LoRA on official Anima 2B (or another trainer-supported official chassis), not on the Delta Mix bake.
Load it on the 3.8B graph as a normal DiT LoRA.
If the LoRA is a 28-block map on a 52-block chassis, use Lora3.8B_only52 from the Delta Mix pack with skip_gen2=true. A raw 28-LoRA on 52 without remap often does nothing.
Do not assume Q35 text-encoder LoRA. This page is Qwen 0.6B + Qwen 3.5 4B + connector. TE keys are not the official 0.6B-only stack.
VRAM: 3.8/52 is heavier than 2B. 12 GB is already tight for inference; training 52-block from this file is not something I have verified.
If you only need style/character on this mix — train on official 2B, test at low strength first.
hey , which folders are the optional files saved into? using comfyUI
They go into the normal ComfyUI folders:
ComfyUI-Anima-Delta-Q35.zip → ComfyUI/custom_nodes/ (unzip so the folder sits next to your other node packs), then restart.
Delta mix 3.8B Q4 1.1.json → anywhere you keep workflows. Easiest: ComfyUI/user/default/workflows/, then Load in the menu.
anima_connector_only.safetensors → same folder as the checkpoint (models/diffusion_models/ or models/unet/, including a subfolder like !Anima3.8/ if that is where V4 sits). Sidecar is not a LoRA.
qwen35_4b.safetensors → models/text_encoders/ (or models/clip/ if that is where your other Qwen encoders already are). CLIPLoader / the Q35 Prompt node should see it after refresh.
Also install, not from Optional Files:
Official runtime: https://github.com/GumGum10/comfyui-anima-3-8B → custom_nodes/
LoRA loader: https://github.com/bestluner-create/ComfyUI-Anima-Delta-Q35-LoRA → custom_nodes/
Graph: UNET (this bake) → Attach connector sidecar → Prompt with 0.6B + 4B, use expanded. Do not load a *Q35bundle in this graph.
@bestluner180 Am i forced to rename each file? seeing that their names change just before downloading them?
thanks for answering
@williamswesley10152 Overall, call it whatever is convenient for you. If the connector is not visible, it’s better to name it as in the original. (on the page in the opt files field)
just wont work, costom nodes dont work either lol
Can you be more specific? “Won’t work” and “custom nodes don’t work” can mean a lot of different things.
What are you running — ComfyUI, Forge, or A1111?
This page is ComfyUI only. Forge/A1111 will not load the Q35 graph.
If Comfy:
portable or Desktop?
ComfyUI version / date?
exact error from the console, not only the red node on the canvas
Checklist that usually kills this file:
Install both packs, then restart Comfy
comfyui-anima-3-8B (GumGum10) — required runtime
ComfyUI-Anima-Delta-Q35 from this page / GitHub
Zip must sit as ComfyUI/custom_nodes/ComfyUI-Anima-Delta-Q35/, not a nested extra folder.
Startup log must show 52 blocks, not blocks=28.
Desktop often never installs the 52 pack. Portable is more reliable.
Do not use official Anima38BV2Loader / *Q35bundle on this bake.
Graph is: UNETLoader (DeltaMixV4Q35) → DeltaMix Q35 Attach (anima_connector_only.safetensors) → DeltaMix Q35 Prompt with both CLIP 0.6B + qwen35_4b, expanded output.
Files that must exist separately (this ckpt is DiT-only, ~7 GB):
anima_connector_only.safetensors
qwen35_4b.safetensors
qwen_3_06b_base.safetensors
qwen_image_vae.safetensors
Use the workflow JSON from this page, not a 0.6B Anima graph and not DeltaQ4B.
Latent must be Anima 16ch. EmptyLatentImage 4ch = black / static.
Paste the first traceback from the console (the ImportError / ModuleNotFoundError / missing node name is enough). Then I can tell you which of the five it is.