local-image-gen

name	local-image-gen
description	Generate images locally using ComfyUI (port 8188) on the ASUS GX10. Use when creating illustrations for books, character art, ACG content, or any image generation task. Includes batch generation, theme packs, and upload to catbox.moe.

Local Image Generation via ComfyUI

Infrastructure

• ComfyUI: /home/hjs/ComfyUI | Port: 8188 | Venv: /home/hjs/comfyui_env
• Primary Model: animagine-xl-3.1.safetensors (SDXL, anime-optimized, symlinked from SD WebUI)
• Pending model: Pony Diffusion XL (ProteusV0.4) — not yet downloaded
• Output Dir: /home/hjs/ComfyUI/output/
• Generation Speed: ~67s per image on GB10 unified memory

Start ComfyUI

source /home/hjs/comfyui_env/bin/activate
nohup python3 /home/hjs/ComfyUI/main.py --listen 127.0.0.1 --port 8188 > /tmp/comfyui.log 2>&1 &
# Wait: curl -s http://127.0.0.1:8188/system_stats | python3 -m json.tool | head -3

Health Check

curl -s http://127.0.0.1:8188/system_stats | python3 -c "import sys,json; d=json.load(sys.stdin); print('ComfyUI OK v'+d['system']['comfyui_version'])"

Single Image Generation

source /home/hjs/comfyui_env/bin/activate
python3 /home/hjs/gx10-book-factory/scripts/gen_image.py \
  --prompt "fantasy sorceress, adult, anime style, masterpiece, best quality, 8k" \
  --negative "nsfw, blurry, deformed, low quality, worst quality, child, loli" \
  --output "/home/hjs/gx10-book-factory/images/output.png" \
  --upload   # prints: PUBLIC URL: https://files.catbox.moe/xxxxx.png

Theme Pack / Batch Generation

source /home/hjs/comfyui_env/bin/activate
python3 /home/hjs/gx10-book-factory/scripts/batch_gen.py \
  --theme "fantasy sorceress" \
  --count 20 \
  --output-dir ./packs/sorceress \
  --upload --zip
# Qwen 27B auto-generates character spec, then 20 variations across outfits/settings/moods
# Creates ZIP pack ready for Gumroad upload

Download Pony Diffusion XL (when needed)

nohup wget -c --progress=dot:giga \
  'https://huggingface.co/dataautogpt3/ProteusV0.4/resolve/main/ProteusV0.4.safetensors' \
  -O /home/hjs/ComfyUI/models/checkpoints/ProteusV0.4.safetensors \
  > /tmp/pony_download.log 2>&1 &
tail -f /tmp/pony_download.log

Prompt Engineering (animagine-xl-3.1)

Positive Template

1girl, solo, <appearance>, wearing <outfit>, in <setting>, <mood> expression,
anime style, anime key visual, masterpiece, best quality, highly detailed, 8k,
adult, 18+, mature woman

Negative (always include)

nsfw, worst quality, low quality, blurry, deformed, extra limbs, bad anatomy,
watermark, signature, username, text, child, loli, young girl

KSampler Settings

• Steps: 28 | CFG: 7 | Sampler: euler_ancestral | Scheduler: normal
• Portrait: 832×1216 | Landscape: 1216×832 | Square: 1024×1024

catbox.moe Upload

• Free, no account required, permanent links
• API: https://catbox.moe/user/api.php
• CRITICAL: Must include reqtype=fileupload field or get 412 error
• Built into both gen_image.py and batch_gen.py

ComfyUI API — Known Bugs Fixed

Bug	Fix
ckpt_name: 'checkpoints/model.safetensors' not in list	Use filename only, no subdirectory prefix
catbox 412 No request type given	Add data={"reqtype": "fileupload"} with requests.post

Aider Integration

cd /home/hjs/gx10-book-factory
aider --read /home/hjs/AIDER_CONSULTANT_GUIDE.md scripts/gen_image.py scripts/batch_gen.py
# Then inside Aider:
# /run source /home/hjs/comfyui_env/bin/activate && python3 gen_image.py --prompt "..." --upload