Generate Studio-Quality Photos Programmatically
Imagine building an AI workflow that generates realistic headshots, LinkedIn photos, or workshop scenes on demand—without hiring a photographer or spending hours editing. The new PhotoGPT skill for Hermes AI does exactly that.
What Problem Does This Solve?
Content creators, recruiters, and marketing teams often need realistic photos for profiles, presentations, or social media. Traditional solutions require either hiring photographers or using generic stock images. PhotoGPT bridges this gap by leveraging AI to generate authentic-looking photos of real people—including support for personal trained models (like the Robin model) that capture specific individuals.
The Hermes integration makes it seamless: generate a photo → automatically upload to Mural via Azure Blob storage. No manual handoffs. No API fumbling.
How It Works
The PhotoGPT skill uses an async job-polling pattern:
- Submit a generation request with a prompt (e.g., "Robin in a professional headshot, blue background")
- Receive a jobId immediately
- Poll the endpoint until the image is ready
- Optionally upscale the result for higher quality
- Upload to Mural if needed
Under the hood, it bypasses Cloudflare blocks by using curl via subprocess instead of Python's urllib—a practical workaround documented in the skill.
Getting Started with Hermes
The PhotoGPT skill supports two model types:
Personal Models — Like the Robin model (ID: 019abc8d-e721-7066-80ee-e7286e27d007), trained on real photos. Perfect for consistent character representation across multiple images. Costs 10 credits per image.
Public Models — Free or low-cost alternatives like nanobanana, seedream, or gpt-image-2. Great for experimentation and bulk generation.
Simply authenticate with your PHOTOGPT_API_KEY and PHOTOGPT_MODEL_ROBIN (stored in ~/.hermes/.env), then craft your prompt:
Generate a professional LinkedIn photo of Robin, medium shot, office background, warm lighting
The skill handles credits validation, training status checks, and error recovery automatically. It even integrates with Mural—generate a photo and upload it directly to your team's workspace.
Real-World Prompt Ideas
- Headshots: "Professional headshot, neutral background, studio lighting"
- Social Media: "Casual portrait, outdoor setting, natural lighting"
- Workshop Scenes: "Group of people in a workshop, collaborative atmosphere"
- LinkedIn: "Corporate portrait, business attire, office environment"
Why This Matters for AI Workflows
Hermes AI agents can now autonomously generate visual content, reducing bottlenecks in content pipelines. Imagine a recruitment workflow that auto-generates candidate materials, or a marketing agent that refreshes team photos monthly—all hands-free.
The async job pattern also teaches a practical lesson in building resilient AI integrations: not every API responds instantly, and polling gracefully handles real-world delays.
Next Steps
Check out the full skill documentation and examples in the hermes-skills repo. The PhotoGPT skill is production-ready with comprehensive error handling and Mural integration.
⭐ If this powers your AI workflow, star the repo: https://github.com/RobinBeraud/hermes-skills
