AI Models

I recommend using the Gemini 2.5 Flash Lite model, which demonstrated one of the best balances between quality and Puter resource usage.

In many cases, the free monthly credit of $0.25 provided by Puter.js is enough to process up to ~1,800 images.

Of course, you can experiment with other models and choose the one that works best for your workflow. And do not forget to monitor your Puter resource usage.

Settings

In AI Stock Metadata Generator, you can choose one of the available AI models as well as a stock platform preset. You can also define your own custom limits for titles, descriptions, and keywords.

All settings are automatically saved in your browser using localStorage and restored after page reload.

Metadata Generation

• Supports JPEG, PNG, WebP, and GIF for metadata generation
• EXIF export is available only for JPEG images
• You can upload up to 100 images at once

If you are not satisfied with the generated results, you can select another AI model, choose a different platform preset, or adjust your custom limits.

Clicking ✨ Generate AI Metadata will regenerate metadata for all uploaded images using the current settings.

Clicking ✨ Generate again will regenerate metadata only for the selected image using the current settings.

Copying and Saving Metadata

There are several ways to copy and save metadata.

• Use the ⧉ Copy to clipboard button available for each metadata field.
• To embed metadata into image files, it is recommended to use the free version of IMS AI Studio for stock metadata management or any similar software.
• Save each image individually with embedded EXIF metadata using the 💾 Save with EXIF button.
• Save all images with embedded EXIF metadata at once as a single ZIP archive using the 📦 Download All as ZIP button.

Regardless of the export method (individual files or ZIP archive), image quality does not change. Only EXIF metadata is added.

The exported image file remains byte-for-byte identical to the original except for metadata changes.

A _meta suffix is added to the filename:
Original: image.jpg
With EXIF: image_meta.jpg

To verify EXIF metadata, it is recommended to use the free version of IMS AI Studio for stock metadata management or any similar software.

Token usage statistics have been added to the Puter generator — both a summary total and a detailed breakdown per model.

A second generator using the Groq API Key has also been added, as a fallback if your Puter tokens run out or if you simply prefer Groq.

Which generates better metadata: Llama 4 Scout (17B) — vision via Groq, or gemini-2.5-flash-lite via Puter?

Gemini 2.5 Flash Lite through Puter, without question.

Gemini is a Google model trained on an enormous volume of multimodal data, with very strong image understanding. Flash Lite is further optimized for speed and structured outputs — making it ideal for metadata JSON.

Llama 4 Scout is a capable open model, but vision is a secondary feature for it, not its core specialization. In practice, it understands image details less precisely and tends to generate less accurate keywords.

Groq / Scout is a great free alternative when your Puter credits are exhausted or you'd rather not register with Puter — but as a primary tool, Puter + Gemini delivers higher quality.

Additional Notes

• I decided not to add a separate “Clear” button to keep the interface as simple as possible. Before uploading a new batch of images, simply refresh the page.
• If metadata was not generated for one or more images, or if a Puter error appears, try clicking the Generate again button. In many cases this resolves the issue.
• If generation still does not work, check whether you have reached your available Puter usage limit.