A tiny Windows app that turns phone DNG photos into normal PNGs that look the way your phone showed them.
Drag your DNGs onto the window, click Convert, get PNGs back.
You shot a photo on your phone in RAW. You convert it online. It comes out dark, flat, weirdly colored.
That's because the picture you remember seeing in your gallery isn't what's actually in the raw sensor data — it's the result of a lot of work your phone did the moment you took the shot: stacking multiple exposures, denoising, sharpening, picking colors. The raw file is the unprocessed sensor capture, and on a phone it usually looks much worse than the photo you remember.
This tool just pulls out the version of the picture your phone already made — every phone DNG has it tucked inside — and saves it as PNG. No darkening, no weird colors, no setup.
- Download
dng2png-vX.Y.Z.zipfrom the Releases page and unzip it. - Double-click
dng2png.exe. - First time only — Windows may say "Windows protected your PC." That's normal for any program that isn't signed by a big company. Click More info → Run anyway. You only see it once.
- Drag your
.dngfiles onto the window. Click Convert.
PNGs are saved next to the original DNGs by default. Use the Choose... button to send them somewhere else instead.
- Drag-and-drop window, handles batches
- Phone-photo orientation is preserved (no sideways or mirrored output)
- The UI stays responsive while it works through the queue
- One portable
.exe, around 8 MB — no installer, no extra runtime - If a file fails, the queue keeps going and you get a Copy error button so you can share the exact reason
- Not a raw photo developer like Lightroom or Camera Raw
- Not a non-Windows app (the binary targets Windows 10+)
- Not signed (that's why the SmartScreen warning shows up on first run)
You'll need Rust 1.92 or newer.
cargo build --releaseThe binary lands in target/release/dng2png.exe.
If you want to hack on it, CLAUDE.md has notes on
how the project is laid out and why a few of the more surprising
decisions were made the way they were.
MIT.