Face Recovery 3

Improved face restoration for low-resolution images

Model Overview

Face Recovery 3 is an updated facial reconstruction model designed to restore recognizable facial detail from low-resolution, damaged, or heavily compressed images. When facial information is missing or unreadable, Face Recovery 3 reconstructs plausible features rather than strictly preserving original pixels.

This model is typically combined with another enhancement model. Face Recovery 3 includes both Realistic and Creative variants, allowing users to balance fidelity and recovery strength based on source quality. The updated model produces more natural and realistic faces with improved skin, hair, and facial detail reconstruction while maintaining coherent, usable portrait results. Because these models are reconstructive by design, results may include subtle changes to expressions or fine facial characteristics.

Face Recovery 3 is intended for portraits and people-focused imagery where standard upscaling cannot recover meaningful facial structure.

Key Capabilities

• Reconstructs facial detail from extremely low-resolution or damaged images.

• Produces more realistic and natural-looking faces with improved facial detail recovery.

• Automatically detects and processes multiple faces within a single image.

• Enables usable portrait results where traditional upscaling fails.

Model Pricing

Output MP	1	4	8	16	24	32	40	50	64	100
Credits	1	1	1	1	1	2	2	3	3	5

Getting Started

Endpoint

• https://api.topazlabs.com/image/v1/restore-gen/async

  • model — "Faces"

Model-Specific Parameters

• grain — bool, optional, default false

  • Whether to add grain to the output

• grain_density — decimal 010 to 1.0, optional, default 0.5

  • Intensity of the grain effect (0.0 to 1.0)

• grain_model — string, optional, default 'silver'

  • Grain model: 'silver', 'gaussian', 'grey'

• grain_size — decimal 1.0 to 5.0, optional, default 1.0

  • Size of the grain particles

• grain_strength — decimal 0.0 to 1.0, optional, default 0.5

  • Strength of the grain effect

• iai_packed_metadata — array, optional, default null

  • Face bounding box and score metadata, as packed list or as a JSON-style string.

• input_height — int, optional, default null

  • Input image height in pixels

• input_width — int, optional, default null

  • Input image width in pixels

• output_height — int, optional, default null

  • Output image height in pixels

• output_width — int, optional, default null

  • Output image width in pixels

• reference_uri — string, optional, default null

  • URI for reference image (faces used in face recovery)

• strength — decimal from 0.0 to 1.0, default 1.0

  • Denoise strength

This model works best on smaller images, where the face in the input image is less than 512 x 512 pixels. For larger images, we recommend using Detail Faces.

Examples

Click to view

Click to view