{"access":{"authentication":{"format":"L402 {macaroon}:{proof_of_payment}","header":"Authorization","protocols":["L402","MPP"],"format_mpp":"Payment <base64url_json>"},"endpoint":"https://sats4ai.com/api/l402/restore-face","method":"POST"},"content_type":"api","cover_url":"","description":"Restore blurry, damaged, or AI-generated faces to sharp, natural quality. Uses CodeFormer (NeurIPS 2022, state-of-the-art on CelebA-Test). Adjustable fidelity — balance between quality enhancement and identity preservation. Stable endpoint — model upgrades automatically. 5 sats per image, no API key or signup needed.\n\nSend the image as base64 in the body:\n```\n{ \"image\": \"<base64 or data URI>\", \"fidelity\": 0.5 }\n```","name":"Face Restoration","pricing":[{"tier":"Standard","amount":5,"currency":"SATS"}],"response_body":{"image_url":{"type":"string","description":"URL to the restored image (temporary, download promptly)"}},"quality":{"model":"CodeFormer","paper":"Towards Robust Blind Face Restoration with Codebook Lookup Transformer (NeurIPS 2022)","benchmark":"CelebA-Test","scores":{"FID":32.65,"LPIPS":0.3668},"real_world":{"CelebChild_FID":109.64,"LFW_FID":125.19,"WebPhoto_FID":167.78},"source":"https://paperswithcode.com/task/blind-face-restoration","note":"State-of-the-art on Papers with Code. Controllable fidelity parameter. Stable endpoint — model upgrades behind it as SOTA evolves."}}