Research profile · Desktop application and Video API model set
Topaz Video AI, researched.
Unusually specific documentation of purpose-separated models. Provisional findings. Explore the findings, limitations and 4 cited sources below.
Topaz Video AI is supported for what it claims to do: recovering and upscaling footage you already have, with an unusually specific and purpose-separated model set. What is not supported is any independent measure of how good the output looks, and its own pricing pages disagree with each other.
- Researched scope
- Desktop application and Video API model set
- Best supported for
- Recovering and upscaling footage you already have
- Principal limitation
- No independent quality measurement; the vendor's own pricing figures conflict
- Accepted sources
- 4 · 3 of 5 criteria supported
Best for, and not for
Best supported use
Recovering and upscaling footage you already have. The strongest finding is unusually specific documentation of purpose-separated models.
Where the evidence warns
No independent quality measurement; the vendor's own pricing figures conflict.
What the evidence supports, criterion by criterion
3 of 5 applicable criteria have enough accepted evidence to say something. The rest are listed below them, unscored.
| Criterion | Evidence | What we found |
|---|---|---|
| Workflow and editability | Moderately supported | The model set is documented and purpose-separated: Precision Upscale and Creative Upscale, frame interpolation via Apollo, Apollo Fast, Chronos, Chronos Fast and Aion, plus slow motion and SDR-to-HDR conversion. Apollo is documented for up to 8x slow motion and Aion up to 16x at higher memory cost.Unusually specific vendor documentation; still vendor-sourced, with no accepted independent evaluation of output quality. |
| Output specifications and delivery | Limited evidence | Documented for archival recovery and upscaling to 4K on macOS and Windows, operating on footage you already have rather than generating new material.No published throughput or hardware requirement figures were accepted. |
| Pricing and value | Conflicting evidence | Now sold by subscription rather than the older perpetual licence. The published monthly figures do not agree: the sales page, the licensing documentation and a later check of the sales page each gave a different amount.Three readings of the vendor's own pricing conflict, so we publish none of them. Annual totals agree more closely than monthly instalments. |
Not established
Researched, but the available public evidence does not support a conclusion. These are gaps, not low scores.
- Visual fidelity and aesthetics — The Korean upscaling paper was located a fourth time on 14 September 2026 (DBpia NODE12550931, DOI 10.51467/ASKO.2025.12.21.4.28). No institutional access was available; the full text was not obtained. The abstract names products but not versions or results.
- Temporal stability and artifact control — Same paper, same status: located, not obtained.
Pricing
Published list price, normalised to a five-second generation. Assumes full use of the plan allocation, before tax, no retakes — so real cost per clip you keep is higher.
| Plan | Cost / 5s | Resolution | Audio |
|---|
All Topaz Video AI pricing scenarios and how they were normalised. Prices verified 2026-09-13 and change without notice.
Alternatives
Same category, so genuinely comparable. We do not rank across categories.
Sources
Every finding above traces to one of these. Version match records how closely each source matches Desktop application and Video API model set.
| Source | Type | Version match | Accessed |
|---|---|---|---|
| Topaz Labs | Official documentation | Exact | 2026-09-12 |
| Topaz Labs | Official documentation | Exact | 2026-09-12 |
| Topaz Labs | Official documentation | Exact | 2026-09-12 |
| Topaz Labs | Official documentation | Close | 2026-09-13 |