Gladas' Workflow for Anima — v2 — v3

Gladas’ v3 ComfyUI workflow is for creating Anima images: use a text prompt or source image, choose the upscale option if needed, and save image outputs.

The upscale path places SeedVR2 before Ultimate SD Upscale (USDU), which reuses the models from the initial image.

The Image Saver is described as saving images with full generation metadata in a CivitAI-readable format.

For SeedVR2, place seedvr2_ema_vae_fp16.safetensors in the vae folder and seedvr2_3b_int8_convrot.safetensors in the diffusion_models folder.

Listed model files include 4x-ClearRealityV1_Soft.pth, animaElement7Model_nexusPrismIII.safetensors, Anzhc Face seg 1024 v2 y8n.pt, bbox/Eyeful_v2-Individual.pt, bbox/hand_yolov8s.pt, and qwen_3_06b_base.safetensors.

Other listed model files are qwen_image_vae.safetensors, sam_vit_b_01ec64.pth, seedvr2_3b_int8_convrot.safetensors, seedvr2_ema_vae_fp16.safetensors, and waisnackmixCollection_wsmAnimaBeta.safetensors.

Listed ComfyUI node packs are ComfyMath, comfyui_ultimatesdupscale, comfyui-custom-scripts, comfyui-easycolorcorrector, comfyui-image-saver, comfyui-impact-pack, comfyui-impact-subpack, comfyui-inspire-pack, comfyui-kjnodes, and rgthree-comfy.

The saved defaults reflect settings used for the author’s first 20 sample images; SeedVR2 denoise ranged from 0.40 to 0.70.

If the int8_convrot model does not suit your hardware, the notes suggest trying another SeedVR2 model or removing the SeedVR2 group.

Suggestion · non vérifié

The v3 notes say CFG was left unconnected to Image Saver; check saved metadata if you need CFG included.

Suggestion · non vérifié

Before running, check that your ComfyUI setup has the intended ComfyMath node pack.

Suggestion · non vérifié