Use case Photography

CraterView, Topaz Labs and Clarity AI compared

Twelve photographs enlarged four times by CraterView, Topaz Labs and Clarity AI, six at high ISO denoised by CraterView and Topaz Labs, the same region side by side, and what each costs.

We enlarged real photographs with four models: CraterView Enhance, Topaz Labs Standard V2 and Wonder 3.5, and Clarity AI Crystal. Topaz Labs and Clarity AI are both industry-leading software solutions focused on AI-powered image and video enhancement. What each tool was given is the photograph as it was taken or posted, so there is no sharper original to score against, and the comparison is the one that matters to you: what each result looks like. Further down, six photographs taken at ISO 12800 and 25600 go through CraterView Denoise and two of Topaz Labs’ denoise models, at their own size; Clarity AI offers no denoiser.

A small, soft photograph of an information board, its map of marshes and its paragraphs of text blurred — Part of an information board at Beccles, enlarged four times with `cv-enhance-v3`. WIPE, ZOOM & PAN to see more detail. Photograph by Ashley Dace, CC BY-SA 2.0, via Wikimedia Commons.

The same part of the board enlarged four times by CraterView Enhance, the map's paths and the text sharp and readable — Part of an information board at Beccles, enlarged four times with `cv-enhance-v3`. WIPE, ZOOM & PAN to see more detail. Photograph by Ashley Dace, CC BY-SA 2.0, via Wikimedia Commons.

How to read the figures

Each figure is one small region of one photograph, as each tool returned it, never resized:

The photograph, its own pixels each drawn four times, so you can see what the tools had to work with.
Lanczos upscale: a standard resize, the kind any image editor does. Lanczos is the method: it makes the photograph four times bigger by smoothing between its pixels, and adds nothing. It shows what the photograph holds and nothing more.
Topaz Labs Standard V2, Topaz Labs’ default upscaler.
Topaz Labs Wonder 3.5, Topaz Labs’ newest model for generating detail.
Clarity AI Crystal, Clarity AI’s precision upscaler.
CraterView Enhance, cv-enhance-v3.

Every tool ran with its own default settings.

Text and maps

Six views of a trail map's legend in a grid of six, two across: the photograph with its pixels visible; a soft Lanczos upscale; Topaz Labs Standard V2, whose letters have turned into shapes that are not letters; Topaz Labs Wonder 3.5 and Clarity AI Crystal, sharp but with misspelt words such as Appcoimate; and CraterView Enhance, which reads Approximate times from the Ecology Centre correctly — ZOOM & PAN to see more detail.

The legend of a trail map at Lynn Canyon Park. The photograph holds the words, just: the Lanczos upscale can be read with some effort. Enhance writes them as they are, Approximate times from the Ecology Centre (one way). Wonder and Crystal draw letters just as sharp but misread several words: Appcoimate, Foology Centre, 30 Foot Foot. Topaz Labs Standard V2 sharpens the blur into shapes that are not letters.

Where the print is smaller than the photograph can hold, no tool can read it. On the same board, the road names beside the legend come back invented from all four, Enhance included. A sharp word is not evidence that it was there to be read.

Six views of a paragraph on an information board in a grid of six, two across: the photograph and the Lanczos upscale, both soft; Topaz Labs Standard V2, its letters misshapen; Topaz Labs Wonder 3.5 and CraterView Enhance, both crisp and correctly spelt; Clarity AI Crystal, readable but softer — ZOOM & PAN to see more detail.

A paragraph on the Beccles board above. Here Enhance and Wonder are both crisp and both spelt correctly; Crystal is readable but softer; Standard V2 misshapes the letters. Photograph by Ashley Dace, CC BY-SA 2.0, via Wikimedia Commons.

Six views of a hand-drawn village map in a grid of six, two across, its red roads, black outlines and a yellow numbered circle: the photograph and the Lanczos upscale, soft; Topaz Labs Standard V2, with pale halos along the lines; Topaz Labs Wonder 3.5, Clarity AI Crystal and CraterView Enhance, all clean and close to each other — ZOOM & PAN to see more detail.

A hand-drawn village map at Harpham. Lines and shapes are easier than letters: Wonder, Crystal and Enhance are close to each other here, and Standard V2 leaves pale halos along the lines. Photograph by Ian S, CC BY-SA 2.0, via Wikimedia Commons.

Hair and fur

Six views of a dark shaggy dog's face and fur in a grid of six, two across: the photograph and the Lanczos upscale, soft; Topaz Labs Standard V2, harshly sharpened; Topaz Labs Wonder 3.5 and Clarity AI Crystal, with many crisp, short strands laid across the locks; CraterView Enhance, softer strands that follow the way each lock falls — ZOOM & PAN to see more detail.

Zoom into the fur. Wonder and Crystal draw more strands, and sharper ones, but many are short, even strokes laid across the locks rather than along them. Enhance’s strands are softer, and follow the way each lock falls: they look like fur in a photograph. Standard V2 sharpens hard. Photograph by Megs Harrison on Unsplash.

Six views of a wolf's eyes and muzzle in black and white in a grid of six, two across: the photograph, grainy; the Lanczos upscale; Topaz Labs Standard V2, speckled; Topaz Labs Wonder 3.5 and Clarity AI Crystal, the coat covered in short hatched strokes, Crystal's the coarsest; CraterView Enhance, fur that follows the shape of the face and the coat — ZOOM & PAN to see more detail.

A wolf in black and white, and the same difference. Wonder and Crystal cover the coat in short hatched strokes, much alike everywhere, Crystal’s the coarsest. Enhance’s fur follows the shape of the face and the coat. All three draw the eyes clearly; Standard V2 turns the photograph’s grain into speckle.

Six views of a hiker from behind with a backpack and a red scarf in a grid of six, two across: the photograph, with dark outlines around the hair and the pack, from sharpening done before we had it; the Lanczos upscale with the same outlines; Topaz Labs Standard V2, sharpened; Topaz Labs Wonder 3.5 and CraterView Enhance, clean, with the outlines gone; Clarity AI Crystal, which keeps the dark outlines — ZOOM & PAN to see more detail.

This photograph already has dark outlines around the hair and the pack, from sharpening done before we had it: you can see them in the photograph itself. Wonder and Enhance clean them away and draw the pack’s zips and logo, and Enhance’s loose hair keeps the direction the wind is blowing it; Crystal keeps the outlines. Photograph by Toomas Tartes on Unsplash.

Rock, the Moon and Mars

Six views of a layered cliff above Lake Agnes in a grid of six, two across: the photograph and the Lanczos upscale, blocky; Topaz Labs Standard V2, noisy; Topaz Labs Wonder 3.5 and Clarity AI Crystal, with crisp rock layers and cracks; CraterView Enhance, with the same layers, softer — ZOOM & PAN to see more detail.

A layered cliff above Lake Agnes in the Canadian Rockies. Wonder and Crystal draw crisp layers and cracks; Enhance follows the same layers, more softly. Photograph by Garythai, CC BY-SA 4.0, via Wikimedia Commons.

Six views of a layered Martian rock with small blue-grey spheres on it in a grid of six, two across: the photograph and the Lanczos upscale, soft; Topaz Labs Standard V2, over-sharpened; Topaz Labs Wonder 3.5, covering the whole rock in crinkled texture; Clarity AI Crystal, close to the Lanczos upscale; CraterView Enhance, with some added grain and layering — ZOOM & PAN to see more detail.

A rock on Mars, with the small blue-grey spheres the Opportunity rover found there. Wonder covers the whole rock in crinkled texture that the photograph gives no sign of. Crystal stays close to the Lanczos upscale. Enhance adds some grain and fine layering, less than Wonder, and keeps the spheres round. Photograph by NASA/JPL/Cornell.

Six views of the rings of a camera calibration target on a Mars rover in a grid of six, two across: the photograph and the Lanczos upscale, soft; Topaz Labs Standard V2, with the small dark marks turned to noise; Topaz Labs Wonder 3.5, Clarity AI Crystal and CraterView Enhance, clean rings and edges — ZOOM & PAN to see more detail.

The colour calibration target on the Perseverance rover’s deck. Wonder, Crystal and Enhance draw clean rings and edges; Standard V2 turns the small dark marks into noise. Photograph by NASA/JPL-Caltech/ASU/MSSS/NBI-UCPH.

Six views of the orange face of a rising full moon in a grid of six, two across, all close to each other: the photograph, the Lanczos upscale, Topaz Labs Standard V2, Topaz Labs Wonder 3.5 with a slightly mottled surface, Clarity AI Crystal and CraterView Enhance, none showing more of the Moon's markings than the photograph holds — ZOOM & PAN to see more detail.

A full moon rising. All six are close: the photograph holds little of the Moon’s markings, and no tool finds more. Wonder mottles the surface slightly. Photograph by Astro Backyard.

Lettering and small figures

Six views of the word Canada printed on the space station's robotic arm above Earth in a grid of six, two across: the photograph and the Lanczos upscale, blurred; Topaz Labs Standard V2, the letters broken; Topaz Labs Wonder 3.5, Clarity AI Crystal and CraterView Enhance, the word clean and readable — ZOOM & PAN to see more detail.

The word on the space station’s robotic arm. Wonder, Crystal and Enhance all write it cleanly; Standard V2 breaks the letters. Photograph by NASA.

Six views of a small hiker with a backpack crossing snow in a grid of six, two across: the photograph, with a bright outline around the figure already in the photograph; the Lanczos upscale; Topaz Labs Standard V2, crunchy; Topaz Labs Wonder 3.5, clean but outlined; Clarity AI Crystal, with heavy outlines and smearing; CraterView Enhance, softer, with most of the outline gone and a dark shape at the hiker's side drawn as something hanging from a strap — ZOOM & PAN to see more detail.

A hiker crossing snow on the Plain of Six Glaciers trail, a few dozen pixels tall, with a bright outline already in the photograph. Crystal makes the outline heavier and smears the figure; Wonder is clean but keeps the outline; Enhance drops most of the outline but is softer, and turns a dark shape at the hiker’s side into something hanging from a strap, which the photograph cannot confirm. Photograph by David Wilson, CC BY 2.0, via Wikimedia Commons.

Noise at high ISO

Six photographs taken at ISO 12800 and 25600, each a window of about a megapixel at the original’s own pixels, through three denoisers:

CraterView Denoise, cv-denoise-v1, which redraws the picture without its grain.
Topaz Labs Denoise Extreme, Topaz Labs’ model for heavy noise at very high ISO.
Topaz Labs Denoise Max, Topaz Labs’ generative denoiser.

Clarity AI offers no denoiser. Each tool ran with its own defaults, and every result came back at the photograph’s own size. Each figure shows the same region of the photograph and of each result, each pixel drawn twice.

Four views of a fruit market's price tag reading P.V.P., 25,00 and KILO at ISO 12800, in a grid of four, two across: the photograph, grainy; CraterView Denoise and Topaz Labs Denoise Extreme, clean, every letter as printed; and Topaz Labs Denoise Max, clean, with P.V.P. written as P.ICP. and KILO as MELKS — ZOOM & PAN to see more detail.

Price tags at a fruit market in Funchal, written by hand and stamped. CraterView Denoise and Denoise Extreme clear the grain and keep every letter. Denoise Max rewrites the stamped words: P.V.P. comes back as P.ICP., and KILO as MELKS. Photograph by Dietmar Rabich, CC BY-SA 4.0, via Wikimedia Commons.

Four views of a smiling woman's eyes and nose at a wedding reception, ISO 12800, in a grid of four, two across: the photograph, covered in coloured grain; CraterView Denoise, clean, with darker lashes and smoother, slightly cooler skin; Topaz Labs Denoise Extreme and Denoise Max, clean, her skin and colour as in the photograph — ZOOM & PAN to see more detail.

A guest at a wedding reception. All three clear the grain from her face. Denoise Extreme and Denoise Max keep her skin and colour as the photograph has them; CraterView Denoise draws darker lashes and smoother skin, a little cooler in colour. Photograph by John Trainor, CC BY 2.0, via Wikimedia Commons.

Four views of a lion's eye and the fur around it, ISO 25600, in a grid of four, two across: the photograph, grainy; CraterView Denoise, many sharp hairs drawn and the coat warmer; Topaz Labs Denoise Extreme, the fur smoothed to a painted look; Topaz Labs Denoise Max, natural fur close to the photograph — ZOOM & PAN to see more detail.

A lion in the bush in the last light. Denoise Extreme smooths the fur into a painted look; Denoise Max keeps it natural and close to the photograph; CraterView Denoise draws the most hairs, more than the photograph shows, and warms the coat. Photograph by Dietmar Rabich, CC BY-SA 4.0, via Wikimedia Commons.

Four views of a carved wooden figure in a green coat on a Christmas market's arch at night, ISO 12800, in a grid of four, two across: the photograph, grainy and soft; CraterView Denoise, crisp and clean; Topaz Labs Denoise Extreme, clean but soft; Topaz Labs Denoise Max, crisp, with two eyes and a mouth drawn on the figure's face — ZOOM & PAN to see more detail.

A carved figure on the arch of the Striezelmarkt in Dresden. Denoise Extreme is clean but soft. CraterView Denoise and Denoise Max are crisp, and Max gives the figure a face, two eyes and a mouth, that the photograph does not show. Photograph by Dietmar Rabich, CC BY-SA 4.0, via Wikimedia Commons.

Four views of the shaded side of a black organ key beside an ivory one, ISO 25600, in a grid of four, two across: the photograph, dark and grainy; CraterView Denoise, clean but lightened to a matte brown; Topaz Labs Denoise Extreme and Denoise Max, clean and as dark as the photograph — ZOOM & PAN to see more detail.

The organ in St Stephen’s Cathedral, Vienna. Both Topaz Labs models clear the grain and keep the black keys black. CraterView Denoise lightens their shaded sides to a matte brown: across this window it is about 30 levels brighter on average than the photograph. Photograph by Dietmar Rabich, CC BY-SA 4.0, via Wikimedia Commons.

Four views of a singer's face on a red-lit stage, ISO 25600, in a grid of four, two across: the photograph, grainy under a red haze; CraterView Denoise, the haze taken out, the background nearly black, his hair dark and his face crisply lit; Topaz Labs Denoise Extreme and Denoise Max, clean, keeping the red haze and the photograph's light — ZOOM & PAN to see more detail.

A singer on a red-lit stage in Dülmen. Both Topaz Labs models keep the red haze of the stage lights. CraterView Denoise takes the haze out with the grain: the background goes nearly black, his hair darkens, and his face comes back crisp but lit as it was not. Photograph by Dietmar Rabich, CC BY-SA 4.0, via Wikimedia Commons.

CraterView Denoise draws the cleanest detail and keeps lettering as it was written, but it does not always keep the light: in dark areas and strongly coloured light it can change the tone of the whole picture. Topaz Labs’ two models kept the tone of every photograph here to within a couple of levels.

Where each one stands

CraterView Enhance keeps closest to what the photograph holds. It reads text best where the text is there to be read, keeps maps and lettering clean, and draws hair and fur the way they lie in the photograph.
Topaz Labs Wonder 3.5 draws the most fine detail of the four. On rock it is often convincing; on hair and fur it is short, even strokes that do not always follow the coat. It also draws texture that is not in the photograph, and misreads small words.
Clarity AI Crystal is detailed and the fastest. It keeps outlines and halos already in the photograph, and makes some heavier, and like Wonder it misreads small words.
Topaz Labs Standard V2 sharpens what is there. It leaves halos and turns small lettering into shapes that are not letters.

For noise:

CraterView Denoise draws the cleanest, crispest detail and keeps lettering as written. It also draws detail the photograph does not hold, and can change the light: it lightened dark surfaces and took the coloured haze out of a stage.
Topaz Labs Denoise Max keeps the tone and natural texture, and rewrites small print and draws faces where the photograph has none.
Topaz Labs Denoise Extreme keeps the tone and the lettering, and smooths fine texture.

Price

All of them charge per image, and some more for a larger result. Every size here is the result’s, not the photograph’s. These are the list prices on 10 October 2026, and on 21 October for the denoisers; the enlargements here were from 1 to 17 megapixels, about 6 for most, and the denoised windows about 1.

	Price per image
CraterView Enhance	1 credit, up to 170 megapixels (prices); free on the community queue, up to 13
Topaz Labs Standard V2	1 credit, up to 24 megapixels; $0.08–0.12 a credit
Topaz Labs Wonder 3.5	1 credit up to 8 megapixels, 2 up to 16, 3 up to 24; $0.08–0.12 a credit
Clarity AI Crystal	$0.05 up to about 4.4 megapixels, $0.10 up to 8.6, $0.20 up to 17
CraterView Denoise	1 credit, up to about 1 megapixel at its own pixels or 10.6 tiled (prices); free on the community queue, up to 8.4 tiled
Topaz Labs Denoise Extreme	1 credit up to 24 megapixels, 2 up to 40, 3 up to 64; $0.08–0.12 a credit
Topaz Labs Denoise Max	1 credit up to 16 megapixels, 2 up to 40, 3 up to 50; $0.08–0.12 a credit

How the comparison was made

Real photographs only, of many kinds: people, pets and wildlife, landscapes, maps and text, and the Moon and Mars. Each was sent at the size it was found. Ten of the twelve here were sent whole; the Mars target and the Six Glaciers hiker come from larger originals, and the tools were sent a 768 × 576 window of each at the original’s own pixel size.
Four times, each tool’s defaults. Only the model, the enlargement and the output format were set. Every result came back at exactly four times the photograph’s size, as PNG, so none was compressed on the way back.
Run on 9 and 10 October 2026 through each service’s public API.
Six high-ISO photographs for the denoisers, each sent as a 1152 × 864 window at the original’s own pixels: about a megapixel, as much as CraterView Denoise works at its own pixels without tiling. Only the model and the output format were set, and every result came back at the window’s own size, as PNG. Topaz Labs’ ran through its public API on 21 October 2026; CraterView Denoise ran the same day on the code the service runs, before it was offered through the API.

Try it on your own file

No signup and no card — a workspace opens for your browser the moment you do.

Open Web App