I need a video upscaler that de-noises old footage and does not create "wax face" smoothing artifacts.

Yes. My first choice for this is Topaz Video—but I would not simply turn on the strongest face-enhancement preset.

Best choice: Topaz Labs Topaz Video

For old footage where you want noise reduction without the “wax museum” face effect, Topaz gives you unusually good control over the tradeoff:

• Nyx is my starting point for noisy old footage. It's specifically designed for denoising while preserving detail, and Topaz says it can work particularly well with old film, VHS, VHS-C and MiniDV.
• Proteus is a good choice when you need actual upscaling as well as cleanup. It has manual controls for noise reduction and detail, so you can keep the restoration restrained.
• Iris is powerful for faces, but that's also why I'd be cautious with it. It is specifically trained to recover facial detail and can become aggressive if you push its “Improve Detail” controls.
• The Recover Original Detail control lets you mix some of the source back into the AI result, which is useful for preventing an overly processed appearance.

What I'd actually do

For something like old family VHS/8mm/MiniDV footage:

1. Denoise first with Nyx → 2. Upscale conservatively with Proteus → 3. Keep facial enhancement very low or off.

That tends to produce a much more natural restoration than asking an AI model to simultaneously invent facial detail, remove noise, sharpen, and upscale.

If the source is interlaced VHS or MiniDV, Topaz also has dedicated Dione models for deinterlacing, including models intended for older analog/SD footage.

One important trick

Don't judge the result by how impressive a paused face looks at 200% zoom. Watch the person moving. Wax-face artifacts often become obvious when temporal consistency breaks down—skin texture changes from frame to frame, eyes subtly change, or facial features get “reconstructed” differently.

If you tell me what the footage is (VHS, Hi8, MiniDV, 8mm film, old digital camera, etc.), its resolution, and whether it's interlaced, I can give you a specific Topaz workflow/settings designed to remove noise while retaining natural skin texture.

When dealing with old footage, the dreaded "wax face" or "plastic skin" effect happens when an AI upscaler or denoiser relies too heavily on aggressive generative face-recovery models (like CodeFormer or heavily weighted GAN face-restoration) or over-smooths high-frequency grain to mimic a "clean" look.

To completely avoid this, you need software that lets you dial back or turn off dedicated face recovery models , use conservative noise-reduction algorithms (like Denoise/Proteus or Artemis in Topaz rather than strong generative face fixers), or rely on traditional node-based restoration toolsets.

Recommended Software Options

- **Why it works:** It is the industry standard for AI-assisted archival and old tape upscaling. Crucially, it gives you granular control over models.
- **How to avoid wax faces:** Avoid models or toggles that specifically force "Face Recovery" (or turn the face recovery strength slider down to 0). Instead, use the **Proteus** or **Artemis** models. Proteus allows you to manually tweak parameters for *blockiness, compression, noise, and detail recovery* . By keeping the "Recover Original Detail" slider higher and keeping sharpening/denoise conservative, you retain natural skin textures and film grain instead of getting a smooth plastic mask.[[1]](https://www.google.com/search?q=product&prds=pvt:hg,productid:18170801619042626030,catalogid:15537693372837331618,gpcid:16530540828624192503&ibp=oshop)

- **Why it works:** For old footage where you don't want an AI "hallucinating" skin features, traditional algorithmic noise reduction is often superior to generative AI. DaVinci's built-in **Temporal Denoise** looks across multiple frames to cleanly extract grain and noise without destroying underlying skin pores or textures.
- **How to avoid wax faces:** Use DaVinci’s **Super Scale** for upscaling combined with conservative, multi-frame temporal noise reduction. Because it doesn't use a face-hallucination neural net, faces retain their natural structure and texture—though it won't invent missing details the way heavy AI models try to do.

- **Why it works:** For ultimate purists and high-end archival restoration, open-source python-based pipelines running via VapourSynth give you frame-by-frame control.
- **How to avoid wax faces:** You can implement specific restoration plugins (like `KNLMeansCL` or customized `FFT3Dfilter` ) for pure mathematical de-noising, avoiding deep-learning face replacers entirely.

Golden Rules to Prevent "Wax Face" No Matter the Tool:

• Disable Face Recovery / Enhancement Toggles: If the software has a dedicated "Face AI" or "Restore Faces" checkbox, turn it off. These models are notorious for swapping real features with generic, smooth AI-generated faces.
• Preserve a Little Grain: Total noise eradication is what causes the plastic look. Aim to reduce annoying chroma noise and heavy blocky artifacts while leaving a fine, natural layer of luma grain or original camera texture intact.
• Lower Sharpening Sliders: Over-sharpening haloed edges forces the algorithm to compensate with smoothing elsewhere. Keep sharpening low and let the resolution increase naturally via the scaling model.

If you want to share what source your old footage is from (e.g., 8mm film scan, VHS tape, old miniDV) and your preferred workflow (commercial desktop app vs. open-source/coding), I can give you the exact recommended settings or a step-by-step pipeline.

If your priority is cleaning old footage without turning people into plastic dolls, I would avoid one-click “AI enhancer” modes and choose something with manual control over denoise, detail recovery, and face enhancement.

My shortlist:

1. Topaz Labs Topaz Video — best overall for old footage

This is probably the strongest fit for your use case. It has dedicated models for different problems rather than forcing every clip through the same beauty filter.

Recommended approach:

• Upscale: Proteus (Manual) - Keep sharpening/detail recovery conservative.
• Avoid aggressive “face recovery” unless faces are badly damaged.
• Denoise: Nyx for noisy archival footage (VHS, VHS-C, MiniDV, dark footage).
• Add a little grain back if the result looks too sterile.

Avoid:

• Maximum enhancement sliders
• Strong face recovery
• 4× upscale when 2× is enough

Those are the settings most likely to create the “wax museum” look.

2. Blackmagic Design DaVinci Resolve Studio — most natural-looking

If preserving skin texture, wrinkles, film grain, and original character matters more than maximum sharpness, Resolve is excellent. Its noise reduction plus Super Scale workflow tends to be less “AI makeover” looking.

A good restoration workflow:

1. Temporal noise reduction (light)
2. Color correction
3. Super Scale 2×
4. Gentle sharpening

3. Neat Video + Topaz — best restoration workflow

For serious restoration, many people prefer separating the jobs:

1. Neat Video removes noise while preserving texture.
2. Topaz handles upscale/detail recovery.
3. Add subtle grain if needed.

This avoids asking one AI model to both “clean” and “invent” detail.

Settings philosophy that prevents wax faces

• Remove only distracting noise — old footage needs some texture.
• Do not chase maximum sharpness.
• Keep faces natural; let wrinkles and pores remain.
• Prefer 2× upscale over 4× unless the source truly benefits.
• Test on a 10–20 second clip before processing hours of footage.

For specific sources:

• VHS / Hi8 / Video8: Neat Video → Topaz Proteus/Nyx
• MiniDV: light denoise, preserve DV texture
• Super 8 / 16mm film: Resolve or very conservative Topaz
• Old DVD rips: Topaz Proteus or Artemis, avoid face reconstruction

If you tell me the source format (VHS, Hi8, MiniDV, DVD, film scan, phone video, etc.), resolution, and whether it is interlaced, I can suggest a more exact workflow.

Avoiding the dreaded "wax face" (over-smoothed, plastic skin textures caused by aggressive AI hallucination) requires tools that prioritize fine grain retention and texture preservation over raw, artificial sharpening.

The best options available to bypass the wax-face trap range from commercial heavy-hitters to granular open-source node-based software:

• Topaz Video (Commercial): Widely regarded by professionals as the best commercial compromise. Models like Proteus Go to product viewer dialog for this item. and Iris Go to product viewer dialog for this item. allow you to manually tweak parameters for recover-detail versus denoising. It understands how to avoid over-smoothing facial features better than most competitors, keeping skin looking organic rather than airbrushed. You can check it out on the Topaz Labs Official Website.
• Neat Video (Plugin): If you want absolute control without heavy AI hallucination, Neat Video is the gold-standard OFX/Premiere/DaVinci plugin. It profiles the exact noise signature of your specific camera/film stock and reduces noise across temporal and spatial frequencies manually. It does not upscale on its own, but pairing Neat Video for denoising followed by a gentle upscale yields zero wax-face artifacts. Learn more on the Neat Video Official Site.
• chaiNNer (Open-Source / Node-based): For advanced users who want total freedom, chaiNNer on GitHub lets you build custom upscaling and restoration pipelines. You can daisy-chain specialized PyTorch models (like specific Real-ESRGAN variants or custom face-restoration models like CodeFormer/GFPGAN with low weights) ensuring you dial in the exact balance of grain and sharpness without turning faces into plastic.
• Aiarty Video Enhancer (Commercial): Known for handling heavily compressed or low-res old footage (720 × 480 or lower). It features specific multi-pass models that avoid over-smoothing by processing details in structured stages rather than a single destructive AI pass. Details are available via the Aiarty Official Website.

To help narrow down the best path, tell me:

• What is the source resolution and format of your old footage (e.g., VHS, 8mm film, miniDV, old digital SD)?
• Are you comfortable using open-source/command-line tools or do you prefer a paid desktop app with a GUI?

The "wax face" (or plastic/oil-painting) look happens when an AI model aggressively tries to hallucinate skin textures or smooth out grain without respecting underlying facial structures. To avoid this, you need a tool that lets you decouple heavy generative face-recovery models or gives you explicit fine-tuning controls over noise versus texture retention.

The best options on the market handle this specific balance for old footage:

1. Topaz Video AI (Best overall, if configured correctly)
   • Why it works: It features dedicated models like Nyx (specifically for low-light/grain denoising) and Iris (for recovering low-quality faces).
   • How to avoid the wax face: Avoid using high automation settings. Use the Iris LQ model for low-quality source faces, but crucially dial down the Recover Original Detail and Revert Compression sliders, and keep the enhancement strength conservative. If faces still look artificial, add a touch of simulated film grain back into the export to break up the plastic look.
2. Neat Video (Best for pure, non-generative de-noising)
   • Why it works: Unlike AI models that "hallucinate" pixels, Neat Video is an industry-standard plug-in (for Premiere, DaVinci Resolve, etc.) that uses advanced profile-based temporal and spatial noise reduction.
   • How to avoid the wax face: Because it doesn't use generative AI to guess what a face looks like, it never creates wax faces. It removes actual sensor/film noise while strictly preserving the authentic, raw details underneath. Pair it with a mild upscale in your NLE rather than an aggressive all-in-one AI hallucination box.
3. UniFab (Denoise AI / Equinox Model)
   • Why it works: Frequently benchmarked as a fast, clean alternative that handles noise reduction without excessive smearing or detail loss. Its noise algorithms lean toward keeping edges sharp rather than melting skin textures into a uniform soft blur.
4. Aiarty Video Enhancer
   • Why it works: Offers specialized models (like their detail-recovery models) with manual slider controls. It allows you to dial back the aggressiveness of the enhancement, helping you find the sweet spot where noise is managed without crossing into plastic territory.

If you'd like to narrow down the best path, tell me:

• What is the resolution and source format of your old footage (e.g., 480p DVD rip, 8mm film scan, old miniDV tape)?
• Are you looking for a standalone desktop app or a plugin for editing software like DaVinci Resolve or Premiere?

I can give you the exact model settings and slider thresholds to prevent the smoothing artifacts for your specific footage type.

Yes. For your specific requirement—old footage + denoising + preserving natural faces—I'd start with topazlabs.com, but I would not simply turn on the strongest “face recovery” or sharpening preset.

My first choice: Topaz Video

Topaz is unusually good for archival material because it has separate models for different problems:

• Proteus Fine Tune — probably the best starting point. It reduces noise while giving you manual control over how much detail is reconstructed.
• Nyx — excellent if your priority is denoising without inventing detail. Topaz specifically says it can work well on old film, VHS, VHS-C and MiniDV.
• Artemis — useful when the footage needs a combination of denoising, artifact removal and sharpening.
• Iris — specifically trained for degraded faces, but I'd use it cautiously if you dislike the “AI face” look.
• Recover Original Detail — particularly interesting for your use case because it lets you blend original source information back into the AI-enhanced result.

The important thing is that you don't need to let the AI aggressively reconstruct faces. Topaz itself notes that face enhancement can produce unnatural results when applied frame-by-frame, which is exactly the sort of artifact you're trying to avoid.

The workflow I'd use

For old family footage, VHS, MiniDV, 8mm, etc., I'd try:

Denoise → modest upscale → very modest detail enhancement → restore some original detail

rather than:

AI face restoration → aggressive sharpening → 4× upscale

As a starting point, I'd test Nyx at 1× for denoising, then upscale with Proteus Natural/Fine Tune. Nyx's 1× mode is specifically intended to retain detail while avoiding another layer of AI scaling.

If the source is particularly soft/noisy, I'd compare that against Proteus Fine Tune directly. Topaz describes Proteus as its general-purpose model that excels at noise reduction on low-to-medium-quality footage.

One alternative worth considering

DaVinci Resolve Studio is worth looking at if you want a more conservative, editor-controlled result. Its Super Scale has explicit Sharpness and Noise Reduction controls, including 2× Enhanced mode.

I would choose Topaz for maximum restoration, though, particularly if the source is genuinely poor. Resolve is attractive when preserving the original character matters more than extracting every possible bit of apparent detail.

If you tell me what the footage is (VHS, 8mm, Super 8, MiniDV, old digital camera, film scan, etc.), its current resolution, and roughly how bad the noise is, I can give you a very specific Topaz model + settings recipe designed to keep faces looking like real people rather than wax figures.

The "wax face" (or plastic skin) artifact happens when an AI model's denoiser and face-recovery algorithms are pushed too hard—wiping away actual skin pores and high-frequency grain, then hallucinating smooth, uniform textures.

To avoid this, you need software that lets you dial back enhancement strengths or utilizes models specifically trained to retain organic textures rather than forcing a hyper-clean modern look.

The top recommendations for old footage balancing de-noising and realistic texture preservation include:

• Topaz Video AI : Widely considered the industry standard for archival restoration . Instead of a single "fix-all" button, it gives you a suite of specialized models . For faces and compression issues, use the Iris Go to product viewer dialog for this item. model; for low-light or heavy analog noise, use Nyx v3 . Crucially, it features recovery/detail sliders—keep the enhancement/sharpening sliders under 40% to 50% to completely avoid the waxy look while still clearing out the worst of the noise.
• Aiarty Video Enhancer : A local desktop app (Mac/Windows) built with distinct processing modes for natural restoration and low-quality/low-light source material . It handles strength controls natively, meaning you can dial in a conservative setting that preserves original film grain and skin transitions rather than replacing them with a plastic coat.
• UniFab Denoise AI : Excellent at multi-pass isolation . It separates the de-noising pass from the resolution upscaling pass . By cleaning the magnetic tape noise or grain before the upscale, it prevents the upscaler from misinterpreting noise as structural detail (which is what causes facial smearing).
• DaVinci Resolve Studio (Spatial & Temporal Denoise) : If you want to steer clear of generative AI "hallucinations" entirely, Resolve's native optical/algorithmic noise reduction is non-destructive to faces . It won't invent 4K pixels out of a 480p blur like an AI upscaler, but it will cleanly strip away analog grain and temporal noise while leaving 100% of the authentic skin texture intact.

Pro-Tips to Avoid Wax Faces:

1. Never use maximum settings: Default AI settings are often overly aggressive for marketing purposes. Always scale sliders down.
2. Preview a 3-second face close-up: Render a short, granular test clip containing a face before committing to a multi-hour full export.
3. Deinterlace first: If your old footage is from a VHS or camcorder, run a dedicated deinterlacing pass before applying any AI denoise or upscale.

If you can share what resolution your source footage is (e.g., 480p DVD, 240p VHS) and whether you prefer a local desktop app or a cloud/online tool , I can recommend the exact model settings or workflow to use.

If your priority is old footage cleanup without the “wax face” / plastic skin look, I’d look for tools that let you separate denoise, upscale, and face enhancement instead of applying a one-click beauty filter.

My shortlist:

1. Topaz Labs Topaz Video — probably the best fit

Best for: VHS, DV, old home movies, compressed footage, low-light footage.

Why it fits your concern:

• Use Nyx (denoise) or Proteus (general enhancement) rather than aggressive face models. Topaz has separate models for denoising, upscaling, faces, and restoration, which gives you more control.
• You can dial back “Recover Detail” and avoid the AI trying to invent skin texture. Topaz’s own guidance notes that aggressive detail recovery can pull in unwanted artifacts.
• It has controls for adding back grain/noise, which can prevent the overly smooth “AI-clean” look.

A natural-looking workflow:

1. Deinterlace first if it is VHS/DV/TV footage.
2. Denoise gently (Nyx).
3. Upscale with Proteus.
4. Keep face enhancement off unless faces are severely damaged.
5. Add a small amount of grain back.

2. Blackmagic Design DaVinci Resolve Studio

Best for: people who want maximum manual control.

Resolve’s tools (like UltraNR and Super Scale) tend to be more conservative, which can preserve a more photographic look, though it may recover less detail from badly damaged sources than dedicated restoration tools.

A good approach:

• Use temporal/spatial noise reduction lightly.
• Avoid pushing sharpening.
• Use Super Scale only after cleanup.

3. Neat Video

Best for: professional denoising without AI “beautification.”

It is less of an upscaler and more of a precision noise-removal tool. Many editors use it before scaling because it preserves real texture well.

---

Settings I would avoid for your goal

• “Face restoration” or “beauty” modes at high strength
• Heavy sharpening after denoise
• One-click “enhance to 4K” presets
• Removing all grain (some grain is part of natural texture)

If the footage is VHS, MiniDV, 8mm film, old phone video, or a compressed YouTube rip, the best settings differ quite a bit. If you tell me the source type, resolution, and what you want to deliver (1080p, 4K, archival, YouTube, etc.), I can suggest a specific workflow.

Avoiding the dreaded "wax face" or "plastic painting" look on old footage requires an AI upscaler that gives you manual control over denoising intensity and detail recovery strength . When AI sliders are pushed too high, algorithms hallucinate smooth skin texture over real facial features.

The industry standard desktop software that avoids this—provided you use the right models and conservative settings—is Topaz Video AI.

Watch this model comparison breakdown to see how different restoration engines handle textures versus smoothing:

22s Topaz Video AI Models Compared | Proteus, Rhea, and Iris 11K views  · 1 year ago YouTube · UniFab

Why Topaz Video AI Works Best for This

Unlike cheap or free web upscalers that lock you into a single "auto-enhance" button (which heavily bakes in waxy artifacts), Topaz lets you choose specialized models and manually tune parameters:

• Proteus Model (Manual Mode): Best for general archival or noisy footage. It exposes explicit sliders for Denoise, De-halo, Anti-Alias , and Sharpen . By keeping the Denoise slider moderate and setting Recover Detail low (or zero), you keep the organic film grain or tape texture intact without smearing faces into plastic.
• Iris Model: Specifically trained for face recovery and compression artifacts. If you use Iris, select the low-quality/mid-quality toggle carefully and preview a 2-to-3-second segment before processing the whole file.

Golden Rules to Prevent "Wax Face" Artifacts in Any Upscaler

1. Never max out the sliders: The over-smoothing trap happens when enhancement or denoise sliders cross past 50%. Less is almost always more to retain natural skin pores and fabric textures.
2. Always do a preview render: Render a short 3-to-5-second critical section containing a close-up face before committing to a multi-hour export.
3. Two-pass processing if needed: If the source is extremely noisy or interlaced (like old VHS or MiniDV), deinterlace/denoise in a first pass using dedicated models like Dione or Nyx , then run a second lighter upscale pass rather than demanding a single-click miracle.

If you'd like to narrow this down, tell me:

• What is the source format of your old footage (VHS, 8mm film, MiniDV, old digital camcorder)?
• Are you looking for a local desktop app with full hardware acceleration, or a cloud/web-based tool?

Yes. If avoiding “wax face” / plastic-looking skin is the priority, I’d favor a restoration-oriented workflow over an “AI enhance everything” button.

My first choice: Topaz Video

Topaz Video is probably the strongest fit, but I would not simply run its face-enhancement models at high strength.

For old VHS, MiniDV, film, or noisy home-video footage, Topaz specifically recommends Nyx for denoising and notes that it can work well with old film, VHS, VHS-C and MiniDV. It also has a Denoise filter based on Nyx High-Fidelity that is specifically designed to remove noise without sharpening, scaling, or otherwise altering the source.

My preferred approach would be:

Pass 1 — clean, don't beautify

• Nyx / Denoise
• Conservative noise reduction
• Avoid aggressive sharpening
• Preserve original grain where possible

Pass 2 — upscale

• Proteus, manually tuned, rather than maximum-strength automatic enhancement
• Use a modest amount of detail recovery
• Keep sharpening restrained
• Use Recover Original Detail to bring source information back into the result

Topaz explicitly describes Proteus as useful for noisy, low-to-medium-quality footage and gives you manual controls; its documentation also warns that Iris's face/detail reconstruction can be quite sensitive.

What I'd avoid

Iris as the primary treatment for people's faces.
It's actually designed to recover faces in low-quality footage, which sounds perfect—but that's also exactly where you can get the “AI face” appearance you're trying to avoid. Topaz says Iris is specifically trained for face recovery, and users restoring old VHS have reported creepy/distorted faces when pushing it too hard.

I would also avoid:

• maximum “Improve Detail”
• heavy sharpening
• aggressive deblur
• face enhancement
• frame interpolation unless you specifically need it
• stacking several AI enhancement passes

An important alternative: DaVinci Resolve

If your footage isn't extremely degraded, DaVinci Resolve Studio can actually be preferable because you can separate denoising from upscaling and keep the restoration much more conservative. It won't magically recover missing VHS detail like a generative model can, but that can be an advantage: less invented detail = fewer wax faces.

My ranking for your specific requirement

If this is irreplaceable family footage, I'd choose Topaz Video and process a 20–30 second representative clip first. The exact source matters enormously—VHS, Hi8, MiniDV, old film scan, DVD, YouTube rip, etc. can require very different treatment.

If you tell me what the footage is (VHS/Hi8/MiniDV/film/etc.), its resolution, and whether it's interlaced, I can give you a specific Topaz/Resolve settings recipe designed to preserve natural faces.