ComfyUI Workflow - Anime Action Character (Full Body + Fast Upscale)

This is my personal ComfyUI workflow for generating high-quality Japanese anime / comic style action scenes with consistent character appearance.

Workflow Overview

• Character LoRA driven generation

• Optimized long prompt for cool-colored anime comic aesthetic

• Low CFG + specialized sampler for clean anime results

• Fast 2-stage upscaling pipeline (great for quick iterations)

• Tall portrait resolution perfect for full-body action shots

Models & Nodes Used

• Base Model: anima-base-v1.0.safetensors

• Character LoRA: personal_NewMisterMR_Elena_V... (strength 1.0)

• CLIP: qwen_3.0b_base.safetensors (custom text encoder)

• VAE: qwen_image_vae.safetensors

• Upscaler 1: z_image_turbo_bf16.safetensors (Z-Image-Turbo)

• Upscaler 2: RealESRGAN_x4plus.safetensors

Key Settings

• Resolution: 1024 × 1536 (tall portrait)

• Steps: 25

• CFG Scale: 4.0 (low CFG works great with this anime model)

• Sampler: er_sde

• Scheduler: simple

• Denoise: 1.0

• Upscale chain: Z-Image-Turbo → RealESRGAN x4plus

Positive Prompt Style

Uses a detailed long prompt focused on:

• Cool color palette (dominant blues & cyans)

• Japanese anime comic / cel-shaded style

• Dynamic full-body running action pose

• Post-apocalyptic ruined city with zombies

• Specific outfit: tight navy blue one-piece swimsuit + brown cowboy boots

Why I like this workflow

• Very consistent character thanks to the LoRA

• Fast upscaling (Z-Image-Turbo makes it quick)

• Low CFG + er_sde sampler gives nice clean anime lines

• Great for action scenes and full-body shots