photocrit

A CLI tool that analyzes a directory of photographs using Claude's vision API and categorizes them into three tiers: failed, good, and keeper. Designed for wildlife, macro, street, travel, and landscape photography.

How it works

1. analyze — sends each image to Claude for individual assessment, then runs a comparative pass to rank similar/burst shots against each other. Writes a photocrit-review.json file with every decision and its reasoning.
2. You inspect (and optionally edit) the review file.
3. apply — moves files into _failed/, _good/, and _keepers/ subdirectories and writes a photocrit-report.md with full per-photo narrative analysis.

Categories

Category	Folder	Description
failed	_failed/	Blurry, over/under-exposed, out of focus, technically unusable
good	_good/	Technically sound but not particularly memorable
keeper	_keepers/	Worth saving, sharing, or posting

Installation

Requires Go 1.21+.

git clone https://github.com/mich8bsp/photocrit
cd photocrit
go build -o photocrit .

Or install directly:

go install github.com/mich8bsp/photocrit@latest

Usage

export ANTHROPIC_API_KEY=sk-ant-...

# Step 1: analyze a directory
photocrit analyze ./photos

# Step 2: inspect the review file, override any decisions if needed
# edit ./photos/photocrit-review.json  (set "override": "keeper" etc.)

# Step 3: move files and generate report
photocrit apply ./photos

Flags

Flag	Default	Description
--model	claude-sonnet-4-6	Claude model for analysis
--opus	false	Use claude-opus-4-6 for the comparative pass
--concurrency	5	Parallel API calls during individual analysis
--review-file	<dir>/photocrit-review.json	Custom review file path
--recursive	false	Recurse into subdirectories
--dry-run	false	Preview actions without writing files or moving images

Review file

After analyze, a photocrit-review.json is written to the target directory. Each photo entry looks like:

{
  "filename": "IMG_0041.jpg",
  "category": "keeper",
  "override": null,
  "reasoning": "Sharp image of a kingfisher mid-dive...",
  "technical": "Well exposed, fast shutter frozen motion...",
  "strengths": ["decisive moment", "clean background"],
  "weaknesses": ["minor fringing on feathers"],
  "group_id": "group_003",
  "group_rank": 1
}

Set "override" to "failed", "good", or "keeper" to change a decision before applying.

Requirements

• Anthropic API key with credits
• Go 1.21+
• Supported image formats: JPEG, PNG, TIFF, WebP — plus RAW formats (CR2, NEF, ARW, DNG, ORF, RW2) with best-effort decoding

Note: HEIC files are not currently supported. RAW file support depends on Go's image decoding capabilities and may vary by format.

Limits

• Hard cap of 200 images per run
• All images are downscaled to a maximum of 2048px on their longest edge before being sent to the API