Re-save a grainy photo as a PNG and look at the file size. We took one clean portrait and damaged it three ways on September 3, 2026: adding noise made the file 2.4× larger, blurring it made the file 21% smaller, and heavy JPEG compression left 4 KB of a 291 KB image. All three look like "a bad photo" on a phone screen. Only one of them is actually grain, and the three need opposite fixes.
Grain, blur and compression artifacts are three different defects that look alike and repair differently. Noise is random detail, so it inflates a lossless file; blur removes detail, so it shrinks one. Fix the wrong one and you destroy what is left. Also worth knowing before you follow most guides: Lightroom's and DxO's AI denoise accept RAW files only — if your grainy photo is a phone JPEG, the two most-recommended tools cannot open it. And denoise before you upscale, never after.
Diagnose it before you try to fix grainy photos
We ran a controlled grainy photos test rather than describing one. Starting from a single clean 1,920px portrait we produced three damaged copies with known parameters, keeping the original for comparison:
| Version | What was done | Lossless file size | Signature |
|---|---|---|---|
| Original | — | 291 KB | baseline |
| Noise | gaussian, σ ≈ 46 | 707 KB | 2.4× larger |
| Blur | gaussian, σ = 3.2 | 230 KB | 21% smaller |
| Compression | JPEG quality 8 | 4 KB | 1.3% of the original |
For grainy photos that size column is the cheapest diagnostic there is, and it works because of what each defect is:
- Grain is added information. It is random, so it does not compress. A noisy image carries more entropy than the clean one it came from, and a lossless format has to store all of it.
- Blur is removed information. Neighbouring pixels become similar, which compresses beautifully. Smaller file, less to recover.
- Compression artifacts are what is left after information was thrown away — blocking and colour bleeding at the 8×8 boundaries, which is why the last tile has visible squares.
The practical test to fix grainy photos correctly
Take your grainy photos and a known-good photo of a similar subject, save both as PNG at the same dimensions, and compare sizes. Notably larger means grainy photos — real noise. Notably smaller means softness. Tiny with visible blocking means compression. You now know which of three different grainy photos repairs to run, which is the step nearly every guide skips.
Half the tools recommended for grainy photos cannot open your file
Here is the constraint most articles about grainy photos leave out. Lightroom's AI Denoise and DxO's are built for RAW input only; Topaz handles JPEG and TIFF.
That matters because grainy photos people actually need fixing are rarely RAW:
| Where the photo came from | Format | Lightroom / DxO AI denoise |
|---|---|---|
| Phone camera roll | JPEG / HEIC | ✗ cannot process |
| Screenshot or download | PNG / JPEG | ✗ cannot process |
| Scanned print | JPEG / TIFF | ✗ / TIFF depends |
| Mirrorless or DSLR, RAW enabled | RAW | ✓ |
| Old family photo, already digitised | JPEG | ✗ cannot process |
So if a guide tells you to "just use Lightroom's AI Denoise" on the grainy photos your relative sent over WhatsApp, that advice cannot be followed. Check what your file actually is before choosing a grainy photos tool — it eliminates most of the shortlist immediately.
Grainy photos: denoise first, upscale second — never the reverse
The two operations do opposite things to grain, and the order decides the result.
Upscaling amplifies the grain. An upscaler's job is to invent plausible detail; grain looks exactly like detail to it, so it enlarges and hardens the speckle into a texture that is far harder to remove afterwards.
Denoising removes fine detail along with the grain. Run it on an already-enlarged image and it has more speckle to chew through, so it strips more real texture to get there.
Run denoise on grainy photos at original resolution first, then enlarge the cleaned file. Our own HD photo converter is the enlargement half of that, and it is free to try before you decide whether the file is worth restoring.
When grain is not the real problem
The controlled grainy photos test above makes one more thing obvious: the blur and compression versions are not grainy photos at all, and denoising cannot fix them. There is no grain in them to remove. Running a denoiser on a soft image just makes it softer.
For prints that are faded, creased or physically damaged, none of these grainy photos fixes applies either — that is old photo restoration, a different operation that reconstructs missing content rather than filtering existing content.
What actually causes grainy photos
Grainy photos have two mechanisms, and they are worth telling apart because only one is preventable after the fact:
- High ISO. Raising ISO amplifies the sensor signal and the sensor's own electrical noise with it. Luminance noise reads as gritty speckle; chrominance noise as coloured blotches. This is real noise and denoisers work on it.
- Underexposure recovered in editing. A dark photo brightened in post amplifies the same noise floor. The grain was always there; the edit revealed it. This is why "fix it later" usually looks worse than getting exposure right.
Both produce grainy photos the tools can address. What they cannot address is detail the sensor never captured — which is the honest limit on every claim in this category.
FAQ: how to fix grainy photos
Why are my photos grainy?
Grainy photos usually come from high ISO or a brightened underexposure. Raising ISO amplifies the sensor signal and its electrical noise together, producing luminance speckle and coloured blotches; brightening a dark photo in editing amplifies the same noise floor after the fact. Both are genuine noise that denoising tools can reduce.
How do I tell grainy photos from blur?
Save the photo as a PNG and compare its size to a clean photo of similar dimensions. In our controlled test, adding noise made the file 2.4× larger while blurring made it 21% smaller — noise is random detail that cannot compress, blur is missing detail that compresses well. They look similar on screen and need opposite repairs.
Can Lightroom fix grainy photos from my phone?
Not with AI Denoise. Lightroom's AI Denoise and DxO's equivalent accept RAW files only, and phone camera rolls contain JPEG or HEIC. Topaz processes JPEG and TIFF, and browser tools work on whatever you upload. Check your file format before picking a tool.
For grainy photos, should I denoise or upscale first?
Denoise first, always. Upscaling treats grain as detail and enlarges it into a texture that is much harder to remove afterwards; denoising a file that has already been enlarged strips more real texture to reach the same result. Clean at original resolution, then enlarge.
Will denoising fix a blurry photo?
No. Blur is missing information, not added information — it is not one of the grainy photos cases — there is no grain to remove, so a denoiser only softens it further. Blur needs sharpening or detail reconstruction, and heavy compression damage needs a different repair again. Diagnosing which of the three you have is the step that decides whether any fix works.
Sources and method
- Our controlled grainy photos test, September 3, 2026: one 1,920px portrait we own, resized to 520px and degraded three ways with sharp — gaussian noise at σ ≈ 46, gaussian blur at σ = 3.2, and JPEG quality 8. Lossless PNG sizes: 291 KB original, 707 KB noise, 230 KB blur; the JPEG is 4 KB. The original is kept for comparison, which is what makes the size signatures verifiable rather than an impression.
- RAW-only constraint: Adobe documents Denoise under Lightroom Classic's raw workflow, and DxO's equivalent is likewise a raw-input tool; Topaz Photo AI processes JPEG and TIFF. We could not extract a verbatim format list from Adobe's page, so treat this as the documented behaviour rather than a quoted clause — and check your own file before choosing a tool, which is the actionable half anyway.
- Noise mechanism: luminance versus chrominance noise and its relationship to ISO amplification.
Last updated: September 2026
