zimage — Free AI Image Generation Skill

Turn text into images for free, right from your AI coding assistant.
Works with Claude Code · OpenClaw · Codex · Antigravity · Paperclip

English · 繁體中文 · 日本語 · 한국어 · Español · Deutsch · Français · Italiano

---

Overview

zimage gives your AI assistant the ability to generate images from natural language. It calls Z-Image-Turbo — an open-source 6B-parameter model by Alibaba's Tongyi-MAI team — through ModelScope's free API.

You:  Create an image of a fox reading a book in a library
AI:   Submitting: a fox reading a book in a library
      Task 91204 — waiting for result…
      Saved → fox_library.jpg

How it compares

	zimage	DALL-E 3	Midjourney
Price	Free	$0.04–0.08 / image	From $10/mo
Open source	Apache 2.0	No	No
Setup time	~5 min	Billing required	Discord required

Free tier: 2,000 API calls/day total, 500/day per model. Limits may be dynamically adjusted. (official limits)

---

Setup

1 — Alibaba Cloud account (free)

Register at https://www.alibabacloud.com/campaign/benefits?referral_code=A9242N

You'll need to verify your phone number and add a payment method during registration. Z-Image itself is free and will not charge you, but Alibaba Cloud requires payment info on file for all accounts.

2 — ModelScope account + binding

1. Go to https://modelscope.ai/register?inviteCode=futurizerush&invitorName=futurizerush&login=true&logintype=register → sign up (GitHub login works)
2. Open Settings → Bind Alibaba Cloud Account and link the account from step 1

3 — API token

1. Visit https://modelscope.ai/my/access/token
2. Click Create Your Token
3. Copy it — looks like ms-xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

---

Install

Claude Code

Say this in Claude Code:

Install the zimage skill from https://github.com/FuturizeRush/zimage-skill

Then:

Set my MODELSCOPE_API_KEY environment variable to ms-your-token

Restart Claude Code.

OpenClaw / ClawHub

openclaw skills install zimage
# or
npx clawhub@latest install zimage

export MODELSCOPE_API_KEY="ms-your-token"

Codex / Antigravity / Paperclip / Other

git clone https://github.com/FuturizeRush/zimage-skill.git
cd zimage-skill
pip install Pillow  # optional, for format conversion
export MODELSCOPE_API_KEY="ms-your-token"

python3 imgforge.py "a sunset over the ocean" sunset.jpg

---

Usage

Through your AI assistant

Just ask naturally:

Generate an image of a cozy coffee shop, warm lighting, cinematic
Draw a pixel-art dragon breathing fire
Create a minimalist logo, blue gradients, clean vector style

Direct CLI

# Basic
python3 imgforge.py "a cat astronaut on Mars"

# Custom size (landscape)
python3 imgforge.py "mountain panorama at golden hour" -o panorama.jpg -W 2048 -H 1024

# Machine-readable output
python3 imgforge.py "abstract art" --json
# → {"ok": true, "path": "output.jpg", "task": "91204"}

CLI flags

Flag	Default	Description
prompt	(required)	What to generate
-o / --out	output.jpg	Output path (.jpg, .png, .webp)
-W / --width	1024	Width, 512–2048 px
-H / --height	1024	Height, 512–2048 px
--json	off	JSON output for scripting

Prompt tips

Quality boosters: 4K, cinematic, professional photography, dramatic lighting, shallow depth of field

Style keywords: oil painting, watercolor, anime, pixel art, ink wash, blueprint, isometric

---

Troubleshooting

Problem	Solution
MODELSCOPE_API_KEY is not set	See Setup above
401 Unauthorized	Use modelscope.ai (not .cn). Bind Alibaba Cloud in settings. Regenerate token.
Timeout	API under load — retry in a minute
Content moderation	Rephrase your prompt

---

Technical notes

• Zero required dependencies — uses Python's urllib.request. Pillow is optional (auto-detected for format conversion).
• Supports custom image dimensions from 512×512 to 2048×2048.
• --json flag enables integration with other tools and scripts.
• Image URLs from the API expire in 24h — the script downloads immediately.
• Model: Tongyi-MAI/Z-Image-Turbo (Apache 2.0)
• API: ModelScope International (api-inference.modelscope.ai)

---

License

MIT-0 — do whatever you want with it.