Data as of Oct 14, 2026Based on 313 AI responses
Reviewed by Dimitry Apollonsky
AI summary
Topaz Labs holds a narrow lead over Adobe Photoshop, driven by recommendations for high-fidelity video upscaling and detail sharpening. For photo restoration that preserves natural skin texture, Adobe Photoshop is the usual answer.
Rank and Mentioned in · since Jul 5
| # | Brand | Recommended for | Mentioned in |
|---|---|---|---|
| 1 | offline video upscaling, de-noising, and artifact-free detail restoration | offline video upscaling, de-noising, and artifact-free detail restoration | 34% |
| 2 | photo scratch removal while retaining natural skin texture | photo scratch removal while retaining natural skin texture | 29% |
| 3 | sharpening footage without introducing artificial motion smoothing | sharpening footage without introducing artificial motion smoothing | 24% |
| 4 | open-source super-resolution and restoring real-world degraded video | open-source super-resolution and restoring real-world degraded video | 14% |
| 5 | non-destructive timeline noise reduction and native video upscaling | non-destructive timeline noise reduction and native video upscaling | 12% |
| 6 | API background removal for transparent glass and veils | API background removal for transparent glass and veils | 10% |
| 7 | developer SDKs for real-time video super-resolution upscaling | developer SDKs for real-time video super-resolution upscaling | 10% |
| 8 | quick mobile-based portrait restoration and face enhancement | quick mobile-based portrait restoration and face enhancement | 10% |
| 9 | automated background removal API integrations for detailed edges | automated background removal API integrations for detailed edges | 9% |
| 10 | client-side video player integrations to reduce streaming bandwidth | client-side video player integrations to reduce streaming bandwidth | 9% |
| 11 | 8% | ||
| 12 | 8% | ||
| 13 | 8% | ||
| 14 | 6% | ||
| 15 | 6% |
What’s changing · since July
Since July, Photoroomrising has become the usual answer for background removal APIs handling transparent objects. DaVinci Resolverising and WebSRrising have also gained prominence for artifact-free video upscaling and player-based super-resolution.
Week of Sep 21–27
Widest split: Topaz Labs, 38% on ChatGPT Search and 29% on Google AI Mode.
| Brand | ChatGPT Search | Google AI Mode | Comparison |
|---|---|---|---|
| Topaz Labs | 38% | 29% | |
| Adobe Photoshop | 28% | 30% | |
| Topaz Video AI | 28% | 20% | |
| Real-ESRGAN | 18% | 10% | |
| DaVinci Resolve | 15% | 9% | |
| Photoroom | 8% | 12% | |
| NVIDIA Maxine | 6% | 14% | |
| Remini | 8% | 11% | |
| Remove.bg | 12% | 6% | |
| NVIDIA | 8% | 9% |
45% of citations point at the ranked brands’ own sites. youtube.com, reddit.com are the most-cited sites outside the ranked brands’ own domains.
Market map
Mentioned in · since Jul 5
Leading brands by topic
Google AI ModeOct 13, 2026
When restoring old portraits, many heavy-handed AI tools tend to turn faces into "plastic" or overly smoothed porcelain. To remove physical damage like scratches and tears while keeping authentic skin pores and fine lines intact, you need a tool that either offers selective brush repair or uses a balanced face-restoration model.
Top options that balance structural scratch repair with skin-texture preservation include:
If you'd like to narrow this down, tell me:
ChatGPT SearchOct 13, 2026
ChatGPT SearchOct 14, 2026
ChatGPT SearchOct 12, 2026
ChatGPT SearchSep 18, 2026
Google AI ModeSep 23, 2026
We use a video player. Who offers a client-side AI upscaling integration to save bandwidth?
Our video streams consume too much bandwidth. Who offers AI-based super-resolution upscaling?
For your needs, I’d prioritize a photo restoration tool that removes scratches locally without applying heavy AI face enhancement. The key is to repair the damaged areas while keeping pores, wrinkles, fine lines, and the natural character of the original photograph.
Based on current comparisons, these are the tools I’d shortlist.
(Best for natural texture)
Use the Healing Brush, Clone Stamp, and content-aware tools to remove scratches selectively. You can avoid smoothing undamaged skin and retain control over every repair.
Best for: portraits, valuable family photos, and precise restoration.
Designed for removing scratches, dust, and other damage from old photographs. It is worth considering when you want to repair specific damaged areas rather than beautify faces.
Best for: old scanned prints with surface damage.
Lets you mark the unwanted areas for removal and fill them using surrounding image information. This targeted approach can help preserve the original look outside the repair. Results depend on the scratch's width and what lies underneath it.
Best for: isolated scratches and relatively simple damage.
One important tip: turn off automatic face enhancement and colorization initially. Work on a copy of the original, remove scratches first, and compare the repaired skin at 100% zoom. Photoshop also supports non-destructive repairs on a separate layer.
One question: Are you restoring old family portraits, where scratches cross people's faces, or photos where most scratches are in the background? That will help me narrow this to the best tool for your situation.
The best tools for restoring scratched photographs without giving people that overly smooth, “plastic” skin are generally editors that let you remove scratches locally, preserve the original image texture, and control AI face enhancement separately.
My main recommendation is Adobe Photoshop for precise scratch repair, with Topaz Photo for additional recovery when a damaged photo needs more detail. The key is to repair the damage without applying heavy smoothing to the entire portrait. Adobe specifically recommends using masks and working on isolated defects, while Topaz warns that aggressive face recovery can make already-detailed faces look unnaturally smooth.
(Best overall)
Best for: Heavily scratched portraits, old family photos, and precise manual repair.
The Healing Brush, Clone Stamp, and Dust & Scratches filter give you control over where repairs happen. Layers and masks let you protect skin texture while removing damage.
Watch out for: Applying broad blur or large-radius scratch removal across the entire face.
Website:
(Best AI enhancement)
Best for: Faded, blurry, low-resolution photographs that need extra detail after scratches are repaired.
Its Recover models aim to reconstruct realistic texture, and its Grain enhancement can help restore a more photographic appearance after processing.
Watch out for: Face Recovery can make already-detailed faces look plastic. Use it selectively and reduce its strength or skip it when the original face is clear.
Website:
Best for: Film scans, dust spots, and controlled manual retouching.
Its ReTouch tool offers Repair and Clone modes, including manual selection of the source area. This makes it useful when you want to fix specific defects rather than let an AI regenerate a face.
Watch out for: Extensive scratches across complex facial features will require patient manual work.
Website:
Best for: Preserving natural-looking skin during portrait retouching.
Its Heal plugin targets skin imperfections while retaining skin texture. It is a useful companion to a restoration editor, particularly for portraits, although it is not my first choice for repairing long scratches across an entire old photograph.
Website:
(Best free option)
Best for: Budget-conscious users comfortable with manual editing.
Its Heal and Clone tools let you replace damaged pixels using nearby image detail, with no need for automatic face regeneration.
Watch out for: Large, complicated scratches can take a lot of manual effort.
Website:
Regardless of the software, this workflow is the most reliable way to get natural results.
1. Repair the scratches first
Use a Healing Brush for small marks and Clone Stamp for long scratches or damaged edges. Work on a separate layer so you can undo or reduce the effect.
2. Preserve pores, wrinkles, and film grain
Avoid aggressive global denoising and smoothing. Use a low-radius Dust & Scratches filter on isolated areas, with masks to protect eyes, hair, and skin detail.
3. Use AI face recovery only if necessary
If the face is genuinely blurred or too small, apply AI recovery gently. Compare against the original at 100% zoom to check that it hasn't invented facial features or erased natural texture.
One important distinction: removing a scratch is different from reconstructing missing facial information. When a scratch runs through an eye, nose, or mouth, AI may invent detail rather than recover the original. For historically accurate family-photo restoration, preserving what remains of the original is often preferable to making the face look artificially perfect.
What kind of photos are you restoring? The best choice depends quite a bit on the damage.
What are you working with?
Old printed family photos
Scanned black-and-white portraits
Faded or blurry digital photos
Professional portrait photographs
What's the main damage?
Scratches across faces
Fine scratches, dust, and speckles
Tears, missing corners, or missing facial details
Blur, fading, and loss of detail
Budget preference
Free tools only
Paid tools are fine
Professional-quality results regardless of cost
[Find the best fit]
Google AI ModeOct 14, 2026
Finding a tool that cleans up heavy scratches without turning a face into a smooth, waxy, plastic mannequin is the holy grail of photo restoration. The "plastic look" usually happens when aggressive AI face-enhancement models hallucinate smooth skin texture or when global noise-reduction/blur filters wipe out pores and fine lines.
The best tools approach this by decoupling scratch/dust removal from face recovery , giving you granular control over the strength of skin texturing.
Why it avoids the plastic look: Topaz separates its processing into distinct, stackable layers rather than applying a single blind "magic" button.[](https://www.youtube.com/watch?v=rm7GuG5TB7E)[[2](https://www.topazlabs.com/see-it-in-action/face-recovery)](https://www.topazlabs.com/see-it-in-action/face-recovery)[[3](https://community.topazlabs.com/t/topaz-photo-ai-4-0-new-dust-scratch-ai-model/90709)](https://community.topazlabs.com/t/topaz-photo-ai-4-0-new-dust-scratch-ai-model/90709)
For your specific goal—removing scratches, dust, and surface damage from old photographs without giving people artificial, overly smooth skin—I’d start with Topaz Photo or ON1 Restore AI. Both are designed for photo restoration, rather than just beautifying faces.
(Best for surface damage)
Its dedicated Dust & Scratch AI model is built to remove physical damage from scanned prints and archival photos, with an updated model focused on preserving original image detail.
Why I recommend it: Start with damage removal before applying any facial enhancement or sharpening.
Official website and documentation
(Best all-in-one restoration)
Its Restore AI tool targets dust, scratches, wrinkles, tears, fading, and color shifts. It also offers prompt-based refinement.
Why I recommend it: Useful if your photos need both physical repairs and careful correction of faded tones.
(Best for beginners)
This Windows program offers tools for removing scratches, creases, stains, and tears, and emphasizes retaining original facial features. It also works offline.
Why I recommend it: A straightforward option for restoring family albums without learning a complex professional editor.
Official website and free download
Whichever tool you choose, use this workflow:
One additional tool worth knowing is Retouch4me Heal. It specifically emphasizes retaining natural skin texture, although it's intended primarily for minor skin defects rather than repairing scratches across an entire old photograph. retouch4.me Try the official demo.
My verdict: Start with Topaz Photo's Dust & Scratch model if surface damage is your main concern. Test it on a copy of one photograph before processing your entire collection.
One question that would help me narrow this down: Are you restoring mostly black-and-white family portraits, or color photographs? That makes a difference when choosing a tool that repairs damage without reconstructing people's faces unnaturally.
Google AI ModeOct 12, 2026
When restoring old portraits, many heavy-handed AI tools suffer from the "plastic skin" effect—completely wiping away natural pores, wrinkles, and authentic texture in favor of an overly smoothed, synthetic look.
The best options that balance aggressive surface damage repair (scratches, dust, and tears) while retaining or realistically reconstructing natural skin texture include:
If glass, veils, sheer fabric, hair, and other semi-transparent edges are important, I’d specifically look for an API that outputs a soft alpha matte, rather than a binary segmentation mask. That’s what prevents the “cut out with scissors” / jagged-edge look.
A few APIs worth testing:
soft edge mode that preserves partial transparency.For your specific requirement, I'd run the same 20–50 difficult images through 2–3 candidates, including:
Then composite the results over black, white, and a saturated color. A mask can look excellent over white but reveal ugly halos when placed over black.
My shortlist for this particular problem would be Runflow + Knockout + PixLab, because their current documentation explicitly addresses transparent/semi-transparent subjects or soft alpha/edge matting rather than merely claiming generic background removal.
If you tell me your volume (images/day), max resolution, language/runtime (Python/Node/etc.), and approximate budget per 1,000 images, I can narrow this down and give you the API call/code for the strongest candidates.
Google AI ModeSep 18, 2026
Handling transparent objects like glass, wine bottles, and veils without jagged edges requires advanced alpha matting and edge-aware segmentation . Standard hard-thresholding APIs will clip semi-transparent pixels, resulting in harsh, pixelated, or haloed edges.[](https://www.edenai.co/post/best-background-removal-apis)
The top API recommendations specifically noted for handling glass, veils, and complex edge matting gracefully include:
ChatGPT SearchOct 6, 2026
Yes. For glass, veils, sheer fabric, hair, and other partially transparent subjects, I would avoid APIs that only produce a hard segmentation mask. You want a soft, per-pixel alpha matte so that pixels can be partially opaque and edges remain smooth.
| API | Transparent objects | Alpha matte | Edge quality | My take |
|---|---|---|---|---|
| Pixelcut | Explicitly claims transparent objects | Yes | Strong |
I am looking for a background removal API capable of processing transparent objects without leaving jagged edges. What are my options?
Finding an AI video upscaler that sharpens details without introducing "hallucinations" (inventing fake textures like plastic skin or distorted text) or shimmering artifacts is tough because aggressive generative models love to over-cook frames.
To avoid the plastic/hallucinatory look, you want tools that rely on traditional or conservative, physics-aware restoration models rather than wild diffusion-heavy "creative" AI.
Here are the top options that prioritize clean, artifact-free precision:
To help you get a better idea of how these perform in motion and handle sharpening without breaking down, here is a breakdown of tools in action:
Watch this overview showcasing various AI upscalers to see how different models handle sharpness and edge retention:
To help narrow this down, tell me:
ChatGPT SearchSep 23, 2026
If your priority is “make it sharper, but don’t invent fake texture or add AI hallucinations”, avoid the most aggressive generative enhancement modes. No upscaler can truly recover detail that was never captured, but some are much better at fidelity-preserving enhancement than others.
My shortlist:
Best when you want:
Use these settings/models conservatively:
Can you recommend a video upscaling tool that cleans up noise in vintage clips without making faces look like plastic or waxy?
I'm looking for software to upscale old video footage. It needs to handle noise reduction effectively while keeping facial textures natural instead of smoothing them out.
ChatGPT SearchSep 26, 2026
For old footage where you want noise reduction + upscaling without the “wax face” look, I’d avoid one-click “enhance faces” tools and use something with manual control over denoise, detail recovery, and sharpening. My shortlist:
Topaz Video has separate models for different problems rather than applying a generic beauty filter. Its Nyx model is aimed at denoising while preserving detail, and Proteus gives you manual control over enhancement strength.
A natural-looking workflow:
VHS / Hi8 / MiniDV / old camcorder:
Start with Nyx for cleanup.
Use Proteus (Manual) for upscale.
Keep detail recovery and sharpening moderate.
Faces:
Google AI ModeSep 26, 2026
Why it avoids the plastic look: Photoshop’s Photo Restoration Neural Filter includes dedicated sliders for Scratch Reduction and Enhance Face . Out of the box, pushing both to 1 0 0% will give you an ultra-plastic face. However, Photoshop is the best for a hybrid manual/AI approach.[](https://www.youtube.com/watch?v=yaaILl8Zpeg)[[2](https://www.topazlabs.com/see-it-in-action/face-recovery)](https://www.youtube.com/watch?v=j2xJblsZUAg&t=2)
Why it avoids the plastic look: Fully automated cloud web-apps (like MyHeritage or random online "AI Photo Fixers") are the worst offenders for plastic skin because they use heavy generative adversarial networks (GANs) trained to output idealized, pore-less faces. Software that lets you dial back AI or isolate regions is far superior.
If you'd like, let me know:
I can give you a step-by-step workflow tailored to the software you prefer.
If you'd like to narrow this down, tell me:
I can give you a step-by-step workflow for the best match.
.png or .webp output to preserve the multi-level alpha channel (semi-transparency). Requesting a JPEG will force a solid color background, flattening transparency.[](https://www.photoroom.com/api/remove-background)channels=alpha ), but note that final edge refinement/matting is often only applied when the API renders the composite cutout directly.[](https://docs.photoroom.com/remove-background-api-basic-plan/download-the-segmentation-mask)If you'd like, share:
I can help recommend the most cost-effective plan or provide a sample API request structure.
| Best mainstream option to test first |
| Adobe Firefly / Photoshop API | Handles complex foregrounds; supports mask output | Yes | Strong | Best if you're already in Adobe's ecosystem |
| remove.bg | Good general-purpose cutouts | Yes | Strong | Very mature/easy, but I'd benchmark it specifically on glass/veils |
| Cutout.Pro | Matting API rather than merely segmentation | Yes | Good | Worth testing for difficult subjects |
| withoutBG | Focuses on alpha mattes | Yes | Good | Interesting option, especially if self-hosting matters |
| Photoroom | Explicitly says it handles transparency/complex edges | Output supports transparency | Strong | Excellent for product imagery |
Pixelcut specifically advertises transparent-object detection, along with hair/fur and intricate boundaries. Its API can return a transparent PNG, an alpha mask, or a foreground + alpha-mask ZIP, and supports images up to 6000×6000.
The alpha-mask output is particularly important for your use case: you can composite the original RGB pixels yourself rather than accepting an already-flattened cutout.
Adobe's current Remove Background API has a mode that can return either a cutout or a grayscale mask, plus a colorDecontamination parameter designed to reduce background color bleeding around edges.
That makes it interesting for veils and glass where edge contamination can be as problematic as the actual segmentation.
This is probably the safest mature API choice. Its ZIP output explicitly contains an alpha matte plus a color image with edge-color corrections. It supports up to 50 MP input.
One caveat: I'd want to benchmark it against Pixelcut/Adobe on actual transparent subjects, rather than assuming its excellent general-purpose reputation translates directly to veils and glass.
Cutout.Pro exposes a matting API and describes the result as foreground colors plus an alpha matte, which is much closer to what you're looking for than a simple binary segmentation API.
This one is particularly interesting technically. Its API has a dedicated Alpha Matte endpoint rather than making alpha an incidental feature of background removal. It also offers self-hostable/open-weight options.
Knockout explicitly demonstrates glass, ice, and liquid where the resulting matte preserves real opacity rather than simply outlining the object. Their site describes the result as per-pixel alpha with what's behind the glass remaining visible.
That's unusually close to your exact requirement, although I'd run your own benchmark before committing to it.
If glass and veils are genuinely critical, my testing order would be:
Knockout → Pixelcut → Adobe → remove.bg → Cutout.Pro
And I would test them using the same 20–50 difficult images, including:
The key metric shouldn't just be "does the background disappear?" You want alpha-matte quality: smooth fractional alpha transitions, preservation of internal transparency, minimal halos/color spill, and no jagged 1-pixel contours.
If you can provide 2–3 representative images (especially one glass and one veil), I can also research which of these APIs/model families is most likely to perform well on those exact cases and compare pricing, latency, resolution limits, and API integration.
A good conservative workflow:
Best when:
Resolve’s Super Scale tends to be more restrained than many AI enhancers. It may recover less dramatic detail than Topaz, but it often keeps a more photographic look.
If your footage is valuable (archive, documentary, family videos), consider:
This will not create fake detail, but it also will not magically restore lost information.
Those are the settings most likely to produce:
| Footage | Choice |
|---|---|
| Clean 1080p camera footage → 4K | DaVinci Resolve Super Scale |
| Old VHS/DVD/compressed footage | Topaz Video (Proteus/Artemis/Nyx) |
| Faces/interviews | Topaz Iris cautiously |
| Film grain/noisy night footage | Topaz Nyx + mild upscale |
| Scientific/archive footage where accuracy matters | Conservative non-AI upscale |
If you tell me the source (e.g., YouTube download, VHS, phone video, anime, movie rip, security footage, 1080p→4K, etc.) and your GPU/OS, I can suggest a specific workflow and settings.
Be cautious with Iris. It is designed for face recovery, which can help badly damaged faces, but pushing it hard is where you are most likely to get that “AI plastic skin” effect.
Turn up Recover Original Detail if the result looks too synthetic; Topaz includes this control specifically to blend back source characteristics.
Add a tiny amount of grain/noise if the footage becomes unnaturally clean.
My starting settings for a family VHS transfer would be roughly:
Resolve Studio gives you a more traditional post-production approach: noise reduction, color correction, grain control, and upscaling in the same environment. It tends to be easier to keep a documentary/archival look because you are not relying as heavily on AI reconstruction.
A common approach:
Neat Video is not primarily an upscaler; it is a very respected noise-reduction plug-in. It is a good choice when you want to remove camera noise while preserving skin texture and fine detail, then handle scaling separately.
If you tell me the source type (VHS, Hi8, MiniDV, 8mm film, DVD, old digital camera, etc.) and target (1080p or 4K), I can suggest a more specific workflow and settings.
Avoiding the dreaded "wax face" (plastic, over-smoothed, or hallucinated facial features) when upscaling old footage requires a tool that gives you manual control over AI generation strength rather than relying on a heavy-handed, black-box model.
The top software recommendations that handle noise reduction gracefully while preserving organic skin textures and film grain include:
To help narrow down the best fit, tell me:
