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
What changed from V2
• 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.
FAQ
Comments (10)
Is Qwen 3.5 bundled in Anima Delta Mix 3.8B Qwen 3.5? I notice the file size is a lot larger than the previous version. If it is baked in would you happen to know if SwarmUI will default to using the baked in text encoder?
No. Qwen 3.5 is not bundled as a working text encoder in this file.
The file is larger because extra tensors were saved with the UNET bake. That attach path does not run. SwarmUI will not get a usable baked Q35 encoder from this checkpoint.
Use the same split as before:
• this file = UNET bake only
• connector sidecar + Qwen 0.6B + qwen35_4b from the normal Anima 3.8B v1.1 / Q35 set
• Comfy graph with the 52-pack nodes
Without the full CLIP stack and those nodes it will not work, same as V2.
@bestluner180 anima_connector_only.safetensors is still necessary to run this model properly, right? So you probably should upload it to the additional files of this model as well. And maybe also Qwen3.5 4B. There is just so much additional stuff necessary to run this model as intended (anima_connector_only file, Qwen3.5 4B, custom nodes by GumGum10, your own custom node), so it would be good if you would provide all of this as easily accessible as possible.
@Lizardon1025 I forgot to add it to the set, thank you for reminding me...
Unfortunately, you’re right — running models from third‑party manufacturers requires a lot of materials. I would be happy to use only the workflow and nodes from the original author myself, but unfortunately, the original nodes expect to receive metadata from that very connector that’s linked to the new 3.8 blocks. If you change them, the bake won’t pass the check, and the nodes simply won’t see it. And if you make a fork, and the node for loading the model sees the new file, Clip Loader itself will no longer accept it… Overall, the model itself is not very amenable to modifications, and there are many subtleties in its operation; it very easily disrupts the output image, and this is specifically due to Q35 and its components.
@bestluner180 Yes I can completely understand how difficult it is for you to build on 3.8B with all its node's limitations. And creating a model that works with this combination of Qwen3 0.6B and Qwen3.5 4B is probably even harder. Even the simple "1girl/1boy" tag seems to be problematic with Qwen3.5 4B as it often understands them as children, but that's less of a problem with your model than with the Anima 3.8B LiquidMix by Liquidn2.
It's generally interesting to see how both Liquidn2 and you try to tackle this Anima 3.8B problem differently. You with the DeltaMix Q35 Attach node and the anima_connector_only file and LiquidMix with the custom model loader (Anima 3.8B LiquidMix + Embedded Semantic Connector v2). Maybe both of you could cooperate in some way to find at least a common solution for this to establish a general preset for all custom Anima 3.8B models.
@Lizardon1025 I talked to LiquidMix; overall, I advised him on what to do with the 2G block and what pitfalls to watch out for, and I also shared a method, but he chose his own path. I considered the option with a built‑in semantic connector, but for that you need to use the Python console for manual stitching; however, this is an inconvenient workaround for those who will be using my method, so it’s easier to do the bake using familiar tools and implement the connector along the way in the workflow. Plus, when you work, you have 1 file of 1.2 GB, not each model with a 1.2 GB bonus weight — this is important when working with dozens of models. In any case, thank you for the praise!
@bestluner180 Yeah I see the advantages of your way as well. It just would be nice if every custom 3.8B Anima model would work in the same way and with the same nodes, but I understand that it is difficult to agree on a certain way to do it.
@Lizardon1025 One more note on the workflow.
It is tested on official Anima 3.8B v1.1 and on this Delta Mix bake: UNETLoader → Attach (sidecar) → Q35 Prompt.
I have not tested LiquidMix. In theory the same graph should load if that file is a normal 3.8 UNET with the 6-block adapter and you use the same CLIP pair (Qwen 0.6B + Qwen 3.5 4B). Attach only adds the official connector weights. I cannot promise the image will match LiquidMix’s own loader.
Thank you for the info. To be honest I only started playing with image generation a couple months ago after I came across a good tutorial for setting up SwarmUI. Setting the right text encoder is easy enough. The rest reads like a different language and it sounds like I’d have to use the comfy tab, which I’ve not yet tried to figure out. I’ve tried hundreds of models and Anima_Delta_Mix_3-8B_Qwen_0-6B_-_DeltaMix3-8B_-_v3-0 was one of the best, so it’d be nice to try Anima Delta Mix 3.8B Qwen 3.5, but it looks like the barrier to entry to too high. I guess I’m waiting for Anima Delta Mix 3.8B Qwen 0.6B V4...
Really impressive! Exceptional prompt adherence—it understands prompt nuances, poses, and complex interactions that I just couldn't get to work in Anima 2B, but Anima 3.8B pulls them off consistently. I definitely can't go back to the standard model. Really looking forward to the int8 convrot release!