Feed a batch of generated images and a reference face. Scores each image in order and stops early the moment one exceeds your target threshold. Returns the best image, score, and how many attempts were used.

AWS Face Compare Batch

⭐ AWS Face Quality (batch)

Score a batch of images for technical face quality — sharpness, brightness, and pose angle. Returns the best image from the batch. No reference face needed.

AWS Face Quality Batch

Example Results

Single compare — same person scoring 98%

Single Compare Result

Batch compare — early exit at attempt 1

Batch Compare Result

Installation

1 — Clone into ComfyUI custom nodes

cd ComfyUI/custom_nodes
git clone https://github.com/Aiconomist/comfyui-rekogniflow

2 — Install dependencies

cd comfyui-rekogniflow
pip install -r requirements.txt

3 — Set up credentials

cp .env.example .env

Open .env and fill in your values:

AWS_ACCESS_KEY_ID     = YOUR_ACCESS_KEY_HERE
AWS_SECRET_ACCESS_KEY = YOUR_SECRET_KEY_HERE
AWS_DEFAULT_REGION    = eu-west-1

or add credentials manually in the custom node.

4 — Restart ComfyUI

Nodes appear under image/face/aws. Leave the aws_access_key and aws_secret_key fields blank in the nodes — credentials load automatically from .env.

AWS Setup (quick summary)

Full walkthrough in the YouTube tutorial.

Create an AWS account at aws.amazon.com
Go to IAM → Users → Create user
Attach policy: AmazonRekognitionReadOnlyAccess
Go to Security credentials → Create access key
Paste both keys into your .env file

Cost: 1,000 free comparisons/month for 12 months (AWS Free Tier), then $0.001 per call.

Node Reference

🔍 AWS Face Compare (single)

| Input | Type | Description | |---|---|---| | image_source | IMAGE | Reference face | | image_target | IMAGE | Image to compare against | | aws_region | STRING | e.g. eu-west-1 — leave blank if using .env | | aws_access_key | STRING | Leave blank — loaded from .env | | aws_secret_key | STRING | Leave blank — loaded from .env | | face_match_threshold | FLOAT | Min score to report a match (default 80.0) | | image_quality | INT | JPEG quality sent to AWS, 50–100 (default 95) |

| Output | Type | Description | |---|---|---| | similarity_score | FLOAT | 0.0 – 100.0 full precision | | detail_json | STRING | Bounding boxes, confidence, full status |

🔁 AWS Face Compare (batch + retry)

| Input | Type | Description | |---|---|---| | image_source | IMAGE | Fixed reference face | | image_batch | IMAGE | Batch of candidates from KSampler | | target_threshold | FLOAT | Stop when this score is reached (default 95.0) | | max_retries | INT | Max images to check, 1–20 (default 5) | | image_quality | INT | JPEG quality sent to AWS (default 95) |

| Output | Type | Description | |---|---|---| | best_image | IMAGE | Winning image | | best_score | FLOAT | Its similarity score | | attempts_used | INT | How many images were checked | | threshold_met | STRING | YES or NO | | detail_json | STRING | Full log of every attempt and score |

⭐ AWS Face Quality (batch)

| Input | Type | Description | |---|---|---| | image_batch | IMAGE | Batch of images to score | | target_threshold | FLOAT | Minimum quality score (default 80.0) | | max_images | INT | Max images to score, 1–20 (default 5) | | weight_sharpness | FLOAT | Sharpness weight (default 35) | | weight_brightness | FLOAT | Brightness weight (default 30) | | weight_pose | FLOAT | Pose/angle weight (default 35) |

| Output | Type | Description | |---|---|---| | best_image | IMAGE | Highest quality image | | best_score | FLOAT | Composite quality score 0–100 | | best_attempt | INT | Which batch image won (1-based) | | threshold_met | STRING | YES or NO | | detail_json | STRING | Per-image: sharpness, brightness, yaw, pitch, roll |

Score Labels

| Range | Label | |---|---| | 95 – 100% | EXCELLENT | | 80 – 94% | GOOD | | 50 – 79% | POSSIBLE | | 0 – 49% | NO MATCH |

License

MIT — free to use and modify. Credit appreciated!
Subscribe → youtube.com/@aiconomist