Wedding photographer, noisy reception
A set of ISO 6400 frames from a dim reception hall. The Topaz Labs denoise pass cleans the shadows while keeping lace detail, and the export prints clean at 12×18.
Process a noisy wedding frameTopaz Labs Examples, Shot by Shot
The examples below are reproducible: a source file in, a processed file out, and the settings that got there. Each scenario covers one job people bring to Topaz Labs — reducing noise, increasing resolution, or repairing damaged scans — and ends with a result you can verify on your own image.
Use cases
What This Scenario Must Deliver
Every set of examples has to answer three questions: what went in, what was run, and what came out.
Family archivist, 1970s scans
Scratched 900px scans with faded contrast and one damaged face. The Topaz Labs restoration stack returns a 2400px master with natural skin tones and no new texture.
Restore an old scanYouTuber, 720p gameplay
Camera-pan footage with grain and compression blocks. The Topaz Labs upscaling pass holds edges during motion, and the final export stays stable at 4K.
Upscale a gameplay clipProduct designer, legacy renders
Small PNG renders from an abandoned tool, used in a client deck. The Topaz Labs sharpen pass adds crisp 2x assets without halos around the product edge.
Sharpen a product renderEditor, archival B-roll
SD footage with heavy artifacts cut into a 1080p timeline. The Topaz Labs video pipeline matches the archival shots to the modern footage in one pass.
Match archival footage
One pair in detail
Each Example, Before and After
Two frames from the examples above, held to the same light.
These topaz labs examples use the same source pair, so the only difference you see comes from the processing models.
Quality bar
The Quality Bar
Set the width of your source clip and see what the Topaz Labs upscaling pass requires on real examples.
3840 × 2160output resolution
15.8 MBestimated file size per frame
24 mintime saved vs manual masking
Rework loop
Common Rework: How the Loop Shortened
Five years of Topaz Labs runs show the same rework loop getting shorter as the models improve.
-
Manual masking
Every artifact had to be painted out by hand; a single example could take three or four passes before it passed inspection.
-
Tuned presets
Denoise and sharpen models ship with defaults that survive a 100% zoom check on the first render.
-
Face recovery
The most common rework — ruined faces — gets a dedicated Topaz Labs model, so the fix becomes a checkbox.
-
Batch queue
Rework becomes the exception. A folder of examples processes overnight and passes inspection in the morning.
-
Default pass
Most source files clear the quality bar on the first export; the loop is down to a single review.
Decision guide
Scenario FAQ: Which Route Fits Your File
The examples above answered “can it be done?” This Topaz Labs split answers which way to go first.
- When the file is noisy but sharp
- Run denoise first and sharpen after. Sharpening a noisy file bakes the grain into edges, so the order decides whether the example passes inspection.
- When the file is soft and undersized
- Upscale, then apply a light sharpen. The upscaler gives the sharpen pass more pixels, and the result holds up at 100% zoom.
- When the file has real damage
- Use the restoration model before changing size. Repairing at native resolution avoids inventing texture that the upscaler would then amplify.
Where does the face-recovery example come from?
The face-recovery example in the archive is built from a 1970s scan with a single damaged face. It shows the Topaz Labs restoration model running before any size change, which is the order the quality bar expects.
What should I compare when I look at an example?
Look at edges, skin texture and background noise. The best examples keep all three stable between the source and the export, rather than trading one for another.
Can I reproduce these examples on my own files?
Yes. Each example lists the source problem and the processing order; the same pass run on a similar file produces a comparable result, though the settings may need a small nudge.
How long does a typical example take?
A single photo example runs in under two minutes on a laptop; a 4K video example takes longer and is usually queued overnight.
The fastest way to build your own example is to run the source file through the same Topaz Labs order: restore, denoise, upscale, sharpen. Start with your own file.