Overview
llmPEG is a lossy image compression experiment that replaces a photograph with a short text description, then reconstructs a new image from that description using a text-to-image model. The project is described as "the LLM Photo Expert Group, after JPEG's Joint Photographic Experts Group." It keeps no pixels and does not preserve faces, text, brands, or the number of objects in a picture. The pipeline uses two local models: a vision model encodes the photo into a text artifact, and an image model later generates a new picture from that artifact. The generator receives only the rendered text, never the source image, and there is no hosted fallback. The project is explicitly not a backup and not JPEG; it is an experiment about how much meaning survives when everything else is deleted.
Pricing
Free
Detailed plans have not been confirmed in our catalog. Check the official website for current limits and billing terms.
Visit WebsitePrices and limits may change. Confirm the currency, billing period, seat minimum and usage caps on the official website.
Use Cases
- Teams use llmPEG to encode a photo into a small text artifact using a local vision model.
- Teams use llmPEG to reconstruct a new image from a text artifact using a local image model.
- Teams use llmPEG to evaluate a reconstructed image against the original with a visual proxy score.
- Teams use llmPEG to verify the format and compatibility of an artifact file.
- Teams use llmPEG to apply tone correction during generation to reduce luminance error.
- Teams use llmPEG to convert a whole folder of photos into artifacts and restored images.
- Teams use llmPEG to experiment with prompt styles that affect reconstruction quality.
Features
AI assistant
Natural language interface
Workflow automation
Reviews
No reviews yet. Be the first to share your experience!