Chatgpt Image Generation

ChatGPT Image Generation Skill

Use Playwright to automate ChatGPT web UI for free image generation (no API costs).

Prerequisites

# Install Playwright
npm install playwright

# Install Chromium browser
npx playwright install chromium

Usage

# Generate images from prompts file
node skills/chatgpt-image-generation/scripts/generate.js --prompts prompts.json --output ./images

# Resume from a specific index
node skills/chatgpt-image-generation/scripts/generate.js --prompts prompts.json --output ./images --start 5

# Use a specific Chrome profile
node skills/chatgpt-image-generation/scripts/generate.js --prompts prompts.json --output ./images --profile "~/Library/Application Support/Google/Chrome/YourProfile"

# Run in headless mode (no browser window)
node skills/chatgpt-image-generation/scripts/generate.js --prompts prompts.json --output ./images --headless

Prompt File Format

["prompt 1", "prompt 2"]

or

{ "prompts": ["prompt 1", "prompt 2"] }

Bulletproof Completion Detection

The detection method checks ALL of the following (must pass):

1. Send button re-enabled — button[data-testid="send-button"] is no longer disabled
2. Image exists — img[alt="Generated image"] present with naturalWidth >= 1024 AND naturalHeight >= 1024
3. Download button visible — Download button appears on hover and is enabled
4. Image src stable — Same src URL detected across 2+ consecutive checks (ensures generation fully done)
5. No progress spinner — No visible [role="progressbar"] or spinner elements

This multi-signal approach prevents:

• Downloading before generation finishes
• Downloading placeholder/thumbnail images
• Missing the download button
• Downloading mid-generation

Download Method

1. Hover over the generated image to reveal the download button
2. Click the download button (uses Playwright expect_download())
3. Save directly to numbered output file

Output

• Numbered image files: 001.png, 002.png, etc.
• results.jsonl — log of success/failure per prompt

Login (One-Time)

If not logged in:

1. Run script once (it will open browser visible)
2. Sign into ChatGPT
3. Session persists for future runs (uses persistent profile)

Platform Support

• macOS: ~/Library/Application Support/Google/Chrome/
• Windows: %LOCALAPPDATA%\Google\Chrome\User Data\
• Linux: ~/.config/google-ch script autorome/

The-detects your OS and uses the appropriate default profile path.

💡 Optional Enhancements

Use a Dedicated Chrome Profile

Benefit: Keep your image generation sessions separate from your normal Chrome browsing. Creates a clean, dedicated environment.

How: Pass --profile with a custom profile path:

--profile ~/Library/Application\ Support/Google/Chrome/ImageGenProfile

Create a new profile in Chrome first, then use its path.

Organize Images in a ChatGPT Project

Benefit: Keep all auto-generated images organized in one project folder in ChatGPT, making it easy to find and review them later.

How: Modify the script to add project selection (see free-image-generation skill as reference). The free-image-generation skill includes this feature by default.

Differences from free-image-generation

This skill:

• ✅ Opens a new chat window (no project selection)
• ✅ Uses default Chrome profile (can specify custom with --profile)

The free-image-generation skill:

• ✅ Automatically selects "Auto Image Generation" project
• ✅ Uses dedicated ImageGenProfile by default