Matthew Choo

Honest reviews and real opinions — no sponsored fluff. I test gadgets, smart home tech, and apps so you know exactly what you’re buying. 20 years in market research means I read the data, not the marketing copy.

2 minute read

Exploring AI-generated images is exciting, but turning it into an automated, repeatable pipeline — especially for stock photo submissions…

Beginner’s Guide to Image Generation Automation: Lessons from My Python Coding Journey

Exploring AI-generated images is exciting, but turning it into an automated, repeatable pipeline — especially for stock photo submissions — requires a mix of creativity, coding, and patience. Here’s what I’ve learned building my own system from scratch, with tips to help beginners avoid common pitfalls.

🔧 Python Coding Tips

  • Start with a menu-based master script  Build your automation script with a clear menu interface that breaks the process into steps. This modular approach simplifies debugging, lets you re-run failed steps, and improves your control over the flow.
  • Make each module independent  Each stage (prompt generation, image creation, metadata embedding) should be able to run on its own. This is crucial for troubleshooting and scalability.

✍️ Prompt Creation Tips

  • Generate in small batches  Prompts created in sets of 10 tend to yield higher quality and more variation. Larger batches may lead to repetitive patterns or degraded creativity.
  • Use online AI models for better results  In my experience, online models like ChatGPT (especially GPT-4 or Claude 3) provide significantly better prompts than local LLMs. If you’re after maximum quality, start online and move offline later.

🧹 Prompt Cleanup Essentials

  • Structured format for metadata  If you plan to embed metadata (like title, description, keywords) into your JPEGs, your prompt structure must follow a predictable format:
Title: "..."
Description: ...
Keywords: ...
  • Flatten your prompts for image generation  ComfyUI’s batch mode works best when you strip prompts into a single-line format. Always save a “flattened” version (one prompt per line) alongside your structured version.

🎨 ComfyUI Setup Tips

  • Command-line installation recommended  Installing ComfyUI via CLI ensures you can run image generation via scripts. This is crucial for full automation.
  • Export correct workflow  Go to Workflow > Export (API) in ComfyUI to get the JSON format compatible with command-line automation.
  • Watch out for ‘text load line from file’ node bugs  In some versions, the index doesn’t reset on repeat executions. Always check your workflow to avoid skipping prompts unintentionally.

🏷️ Embedding Metadata into Images

  • Auto-match metadata with filenames  To automate uploads to sites like Adobe Stock or Dreamstime, match each prompt with the corresponding JPEG (e.g., 001.jpg ↔ 001.txt block).
  • Use IPTC format for metadata  Tools like Pillow and piexif (Python libraries) let you inject IPTC metadata into images. This improves discoverability on stock sites.
  • Prepare differently for each stock platform
  • Dreamstime is more lenient with descriptions but may require FTP-based upload scripts.
  • Adobe Stock is strict — ensure your metadata follows their character limits and keyword best practices.

🤝 Need Help?

Feel free to contact me at mattckw@gmail.com if you’re building something similar. I’m happy to share sample scripts, workflows, or help troubleshoot your setup.

Where to Buy

GPU for AI / Local Image Generation
GPU for AI / Local Image Generation
Run Stable Diffusion locally — minimum 8GB VRAM recommended
Shopee AliExpress
RTX 4060 (8GB) is the sweet spot for local AI image generation at current prices.

Support Me

If this helped, consider buying through the links above — it costs you nothing extra and keeps this blog going.

You may also enjoy