What kinds of face damage can it fix?

Blurriness, pixelation, compression artifacts, noise, and general degradation. Best results on faces that are recognizable but lack detail.

Does it change how people look?

No. GFPGAN preserves identity and natural features. It enhances detail and sharpness without altering facial structure or appearance.

v1.4 is recommended for most use cases — it preserves the most detail and identity. RestoreFormer is an alternative for very damaged footage.

2 to 6 minutes for typical clips. Processing time depends on video length, resolution, and how many faces appear in each frame.

The model focuses specifically on faces. Background and non-face areas receive upscaling but not the specialized face restoration treatment.

2x is the default and recommended setting. This doubles the resolution while keeping faces natural-looking and detailed.

Yes, though it's designed for degraded footage. Modern HD/4K video with clear faces won't benefit much from face restoration.

Each restoration costs 2 credits. Credits are automatically refunded if processing fails for any reason.

AI face restoration (GFPGAN)

GFPGAN v1.4 recovers face sharpness and detail without introducing artifacts. Fidelity slider ranges from natural to aggressive reconstruction.

Continue your workflow with these tools