Topaz Generative

Topaz Generative

Vidu Upscale Pro enhances uploaded or generated videos to 1080p, 2K, 4K, or 8K.

Upload or choose a source video in Models or Generate, or start from a completed video's TaskCard.

Why choose Topaz Generative

Key strengths

• Diffusion-based restoration that synthesizes plausible detail instead of interpolating it.
• Recovers faces, hair, and fabric texture that precision upscaling leaves soft.
• Four engines trading fidelity against sharpness.

Good fit for

• Restoring badly degraded footage where precision upscaling is not enough.
• Hero shots that need to hold up at full size.
• Short clips where quality matters more than turnaround.

Engines in Topaz Generative

This model is a price tier: every engine below costs the same, so the choice is about the source, not the budget. Pick one in Generate under the model; leaving it unset uses the first.

• Starlight HQ - Highest quality of the diffusion models and the slowest. The choice when the result matters more than turnaround.
• Starlight Precise 2.5 - Stays closest to the source; the safest generative option when fidelity to the original matters.
• Starlight Mini - A lighter, faster pass at the same price tier. Good for longer clips or iteration.
• Starlight Sharp - Favours crisp edges over smoothness; suits graphic and high-contrast material.

Inputs and output settings

Upscaling is live in Models, Generate, and on a completed generation's TaskCard.

How to upscale a video

1. Upload a source video or choose one from Assets in Models or Generate. You can also start from a completed video's TaskCard.
2. Select 1080p, 2K, 4K, or 8K above the source resolution.
3. Review the estimate, then submit the job.
4. Inspect the enhanced video when processing completes.

How to get better Topaz Generative results

• Slow and expensive by design - reach for it on shots that justify it, not a whole timeline.
• Clips are capped at one minute; cut the source down to the section that matters.
• Because detail is synthesized, check that faces and text still read as the same person or words.
• Try Generative Fast first if you need to preview the treatment on a longer clip.

Limits and availability

• Choose a target above the source resolution: 1080p, 2K, 4K.
• Maximum source length is 60 seconds.
• Any frame rate is accepted. Billing is per delivered frame, so a 60 FPS source costs twice a 30 FPS one of the same length.
• Direct uploads are limited to 256 MiB; generated MachGen videos can be selected from Assets or a completed TaskCard.
• Current rates come from the live MachGen pricebook.
• Topaz Generative is pass-through. MachGen sends the request through its upstream provider integration and returns the generated asset.

Learn more

For model background, technical details, and architecture, see the official Topaz Starlight documentation.

Use the MachGen Playground to upload or choose a source video. You can also open a completed video's TaskCard and choose Upscale.

Official documentation may describe capabilities or parameters that are not currently exposed by MachGen. Use the Inputs and output settings section above as the source of truth for this page.

Topaz Generative pricing

See Pricing for currently published MachGen rates. Models, Generate, and the completed video's TaskCard show the current estimate before an Upscale job is submitted.

Topaz Generative on MachGen

Deployment: Pass-through. Topaz Generative is pass-through. MachGen sends the request through its upstream provider integration and returns the generated asset.

Upload or choose a source in Models or Generate, or open a completed MachGen generation's TaskCard and choose Upscale.