Photo Style Unifier
Extract the light. Apply the vision.
Upload a few reference photos. Photo Style Unifier reads the colour, tone and atmosphere that runs through them, then applies the same feel to your own photos.
- Free. No account, no limits, no watermark.
- Private. Everything runs in your browser. Your photos are never uploaded anywhere.
- Open source. MIT licence.
- Made by office 未来圏.
👉 Use it here: https://kohyoshida1-dev.github.io/photo-style-unifier/
How to use
- Reference. Drop in 1–8 photos whose look you like (3+ recommended).
- Analyze. The app extracts a style signature: tonal zones, hue-by-hue saturation, light-chroma correlation, softness, grain and more.
- Apply. Drop in your photos, preview live, adjust intensity, and download at full resolution.
Saved styles are kept in your browser's local storage.
Supported formats: JPEG, PNG, WebP (and HEIC in Safari). Convert RAW files to JPEG first.
How it works
The algorithm doesn't copy a LUT or a single average colour. Instead it:
- splits each image into shadow / midtone / highlight zones (with smooth cross-fades) and measures the Lab colour cast in each,
- measures average chroma in 12 hue buckets of 30°,
- takes the median across all reference photos, so one odd photo doesn't skew the result,
- shifts the target's zones towards the reference while protecting skin tones and already-saturated colours,
- adds film-like highlight roll-off, softness, micro-contrast and grain where the references have them.
All processing lives in src/engine/ and runs in a Web Worker.
Development
npm install
npm run dev # http://localhost:5173
npm run build # outputs to dist/
Pushing to main deploys to GitHub Pages automatically (see .github/workflows/deploy.yml).
日本語
参考にしたい写真を数枚選ぶと、その色・トーン・空気感を読み取り、自分の写真に同じ雰囲気を適用できるツールです。
- 無料：会員登録・枚数制限・透かしはありません
- プライバシー：処理はすべてブラウザ内で行われ、写真がどこかへ送信されることはありません
- オープンソース：MIT ライセンス
- 制作：office 未来圏
使い方：上の URL を開き、「お手本の写真」を入れて解析し、加工したい写真を入れて強度を調整し、ダウンロードするだけです。
License
MIT
