Higgsfield image
Generate images using Higgsfield's Nano Banana Pro (nano_banana_2) via the mcporter MCP server. Use for product photos, character-consistent fashion photography, and creative assets. Primary image generation tool for TracyToronto e-commerce content.From its SKILL.md
npx -y skills add S3YED/appie-kit --skill higgsfield-imageAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 6 stars6 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
6.5 KB, ~1.7k tokens by cl100k_base, as published. Nobody here has run it
Higgsfield Image Generation
Auth System (CRITICAL — two separate auth mechanisms)
Higgsfield has TWO auth systems. generate_image requires OAuth tokens, NOT the API key.
| Credential | Type | Used For | Location |
|---|---|---|---|
HIGGSFIELD_API_KEY_ID + HIGGSFIELD_API_KEY_SECRET | API Key pair | tools/list, balance (read-only) | ~/.openclaw/workspace/config/mcporter.json under higgsfield.env |
| OAuth Bearer token (from Clerk) | OAuth session | generate_image, generate_video, show_characters (write ops) | ~/.mcporter/credentials.json under higgsfield|fb7b405dd99c24b1 |
The API key pair does NOT work for generate_image. That tool requires a valid Clerk OAuth access token.
Quick Reference
Tool: mcporter call higgsfield.generate_image
Model: nano_banana_2 (Nano Banana Pro, 4K quality)
Auth: OAuth Bearer token from mcporter credentials
Generate an Image
mcporter call higgsfield.generate_image params='{"model":"nano_banana_2","prompt":"YOUR PROMPT","aspect_ratio":"4:5"}'
With Reference Image (Character Consistency)
mcporter call higgsfield.generate_image params='{"model":"nano_banana_2","prompt":"YOUR PROMPT...same woman as reference...","aspect_ratio":"4:5","medias":[{"value":"https://URL-TO-REFERENCE-IMAGE.PNG","role":"image"}]}'
IMPORTANT: Use value (not url) for the media reference.
OAuth Token Refresh
If generate_image returns "Invalid or expired token", the OAuth session has expired:
mcporter auth higgsfield --reset
This opens a browser for Clerk OAuth login and stores new tokens in ~/.mcporter/credentials.json.
Token status check:
cat ~/.mcporter/credentials.json | python3 -c "
import json,sys
d=json.load(sys.stdin)
e=d['entries'].get('higgsfield|fb7b405dd99c24b1',{})
print('Has tokens:', 'tokens' in e and bool(e.get('tokens')))
print('Updated:', e.get('updatedAt'))
"
Check Generation Status
mcporter call higgsfield.show_generations type="image" size=5
Poll every ~30 seconds. 4K images take 30-50 seconds to generate.
Models Available
| Model ID | Name | Best For |
|---|---|---|
nano_banana_2 | Nano Banana Pro | Top quality 4K, text/diagrams, fashion |
nano_banana_flash | Nano Banana 2 | Fast 4K, high-volume work |
soul_2 | Soul 2 | Character/avatar consistency |
marketing_studio_image | Marketing Studio | Product ads with avatars |
Aspect Ratios
1:1— Square (Instagram post, thumbnail)4:5— Portrait (Instagram post, fashion)2:3— Portrait tall (fashion editorial)3:4— Portrait (standard)9:16— Story/Reel vertical16:9— Landscape (YouTube, banner)21:9— Ultrawide (banner)
Resolution Options
1k— 1024px (fast draft)2k— 2048px (standard)4k— 4096px (premium, default for nano_banana_2)
Character Photo System Prompt
Always include for realistic fashion photography:
The image must look like it was captured with a high-end DSLR fashion camera using an 85mm lens, with natural depth of field, realistic light diffusion, and authentic photographic detail. The result must be extremely photorealistic, studio-quality fashion photography in ultra 4K resolution, indistinguishable from a real camera photograph, with no AI artifacts or synthetic appearance. Maintain true-to-life skin texture, natural wrinkles, pores, and realistic lighting reflections. No beauty filters, no artificial smoothing.
Complete Workflow Example
1. Generate product photo with character
# Start generation
RESULT=$(mcporter call higgsfield.generate_image params='{"model":"nano_banana_2","prompt":"Full-length studio fashion photograph of a mature woman (47 years old, Franco-Canadian, elegant professional style, same face as reference) wearing a cozy plaid fleece jacket in rich autumn colors. She stands confidently in a clean white studio setting, full body visible, natural depth of field, 85mm lens photography, ultra realistic fashion editorial, authentic photographic detail, no AI artifacts.","aspect_ratio":"2:3","medias":[{"value":"https://d8j0ntlcm91z4.cloudfront.net/user_3AFZ6u2cNdsIdJPF7lWsctX8zrH/hf_20260418_172507_97ef24b2-a658-4cd9-acb8-00429c66948d.png","role":"image"}]}')
JOB_ID=$(echo $RESULT | python3 -c "import json,sys; print(json.load(sys.stdin)['results'][0]['id'])")
echo "Job ID: $JOB_ID"
# Wait and check
sleep 40
mcporter call higgsfield.show_generations type="image" size=5 | python3 -c "
import json, sys
data = json.load(sys.stdin)
for g in data.get('items', []):
if g.get('id') == '$JOB_ID':
print('STATUS:', g.get('status'))
if g.get('status') == 'completed':
print('URL:', g.get('results', {}).get('rawUrl'))
"
2. Download generated image
curl -s "IMAGE_URL" -o ./generated_image.png
Quality Notes
- nano_banana_2 produces 8/10 fashion photography — excellent skin textures, realistic lighting
- 4K generation takes 30-50 seconds
- Reference images must be accessible HTTPS URLs (no local files in MCP context)
- For Margot character: use
margot_APPROVED.jpgas reference
Troubleshooting
"Invalid or expired token": OAuth session expired. Run mcporter auth higgsfield --reset on the Mac Mini.
"Model not found": Check model ID is correct — use nano_banana_2 not nano-banana-pro
"Invalid arguments: expected string, received undefined": The medias field requires {value: string, role: string}, not {url: ..., roles: [...]}
Generation pending forever: Poll show_generations with size=5 to find your job ID
Credentials Locations
| File | Content |
|---|---|
~/.mcporter/credentials.json → entries.higgsfield|fb7b405dd99c24b1 | OAuth PKCE data + tokens (state, codeVerifier, access_token, refresh_token) |
~/.openclaw/workspace/config/mcporter.json → higgsfield.env | HIGGSFIELD_API_KEY_ID + HIGGSFIELD_API_KEY_SECRET (API key pair — does NOT auth generate_image) |
The API key pair and OAuth tokens are TWO DIFFERENT things. mcporter.json's API key only works for read-only tools like tools/list.
What ships with it: 1 file
6.4 KB alongside SKILL.md, 1 of them executable
scripts/
- higgsfield_image.pyruns6.4 KB