DEEP ANALOG

This service is in private testing.
Enter your access code to continue.

DEEP ANALOG
machine learning powered color rebuild
Paper GitHub Mechanisms Privacy
Processed in memory · never stored · no tracking

Extract the look of any film and frame.

Upload a film scan or any image you love — the model predicts a 3D LUT that captures its color character, ready for Photoshop, Lightroom, DaVinci Resolve, or any editor. Optionally add one of your own photos to see the full film rendering applied.

●  REFERENCE — required
🎞
Drop a reference image
The film look you want to extract.
JPEG · PNG · TIFF · WebP · HEIC/HEIF · AVIF · RAW (CR2/CR3, NEF, ARW, RAF, DNG, …)
○  YOUR PHOTO — optional
📷
Drop your photo
Add a photo to get it rendered in the reference look — color, tone, grain and halation. Leave empty to just download the LUT.

Rendering controls

Results

See it in action

1  Pick a reference — any film look you love
Reference film still
Still from Happy Together (1997), dir. Wong Kar-wai — shown for demonstration.
2  Add your own photo — the input
Input photo
3  Get it rendered in that look. Drag the divider to compare.
Rendered result Original photo
RENDERED ORIGINAL
loads both images into the app above

Try it yourself — pick a pair

Choose a reference look and an input photo, then run the pair through the model. Film-stock pairs with the seven test photos also show an instant pre-rendered preview. All test photographs shot by Eton Mu.

1  Choose a reference

5 real film stocks, scanned from Eton's negatives. Click a canister for the full story of that emulsion.

2  Choose an input photo
3  Result
Selected reference → Selected input

Rendered Original
RENDERED ORIGINAL

Pre-rendered preview · drag the divider to compare

No pre-rendered preview for this pair.
Run it live — the model predicts the LUT from this reference in seconds.
Download the look
LUT.CUBE Adobe.XMP Capture One.COSTYLE Parameters.JSON
Deep Analog — open-set film emulation · research preview · patent pending.
Created by Eton Mu ↗ · Your images are processed in memory and never stored. How privacy works here.

Privacy by Design

How this service handles your images
📱
Your device
you pick the images
→HTTPS 🔒
⚡
Server RAM only
processed in memory,
discarded instantly
→same response
🎞
Back to you
LUT + rendered photo,
metadata stripped
✗ disk / database ✗ logs of your files ✗ cookies & analytics ✗ third-party requests ✗ EXIF / GPS in outputs

💾 Nothing is stored

Your images are processed entirely in server memory (RAM). They are never written to disk, never saved to any database, and are discarded the moment your result is sent back. There is no way for us to retrieve an image after your request completes.

🗒 Nothing is logged

Server access logs are disabled. We do not record filenames, image contents, IP-address logs, or the results generated for you. Error handling logs only the type of failure, never your data.

🍪 No tracking

This page sets no cookies, runs no analytics, and loads nothing from third-party servers — every byte comes from this app itself. Your access code is kept only in your browser's session storage and vanishes when you close the tab.

📍 Metadata is stripped

Rendered images are re-encoded from raw pixels, which removes all embedded metadata (EXIF, GPS location, camera serial numbers) from anything you download.

🔒 What leaves your device

Only the image files you explicitly upload, over an encrypted HTTPS connection. The predicted LUT and rendered image are returned in the same response and belong to you.

🖥 Hosting

The service runs on Hugging Face infrastructure; transport is encrypted end-to-end. No third party is granted access to request contents by this application.

What happens in RAM, stays in RAM.