Skip to content

Project Starlight

Intermediate

Development line: project:project-starlight · thread project-starlight-public-rollout
Last event: 2025-02-11 · 2 dated since 2025-02-06 · Researched: 2026-09-04 · confidence: high

What it is

Topaz's original generative diffusion model for restorers of archival video and low-resolution GenAI clips.

  • Resolution upscaling.
  • Structural reconstruction.
  • Temporal consistency recovery.

The original model runs in the API as Starlight Precise 1 (sl-1) without manual controls, capped at 9 000 input frames. It is not one fixed product but the start of a model family, so choose by specific Starlight variant and run target.

Development line

  • 2025-02-06 — Topaz Labs publicly referenced Project Starlight. A cloud research diffusion model for video restoration, prioritizing quality over speed and model size.
  • 2025-02-11 — Project Starlight appeared in Topaz Labs Experiments. On 2025-02-11, a Project Starlight URL in Topaz Labs' Experiments web application was recorded. The supplied evidence does not preserve page contents, so this entry records the dated experiments endpoint without claiming a precise workflow, eligibility, or feature set.

What changed

  • 2025-02-06 — Topaz introduced Project Starlight as a cloud research diffusion model for video restoration, prioritizing quality over speed and size.
  • 2025-02-11 — The model became available through an experimental web interface. At the time, free previews and paid renders carried different content terms.
  • 2025-03-05 — Early access added 4K Starlight renders.
  • 2025-03-25 — Topaz prepared the first locally runnable variant of the series.
  • 2025-05-13 — Starlight Mini launched for local rendering on Windows with NVIDIA GPUs.
  • 2025-09-17 — Starlight Sharp expanded the local and cloud lineup with a more detailed variant.

How to use this

Treat the Topaz Labs Experiments route from 2025-02-11 as a dated discovery path. Verify the current experiment page before attempting use, because these links alone do not establish capabilities or access requirements.

  1. Decide whether we need a local model in Topaz Video or a cloud model. The documentation lists Mini, Sharp, Fast 2, HQ, and Precise 2.5/2.6, and notes that cloud models also run in Astra. — https://docs.topazlabs.com/topaz-video/project-starlight-series
  2. For local rendering, install Topaz Video alongside Neuroserver. All Starlight diffusion models require it. If Neuroserver fails, Topaz advises reinstalling the application in Repair mode. — https://docs.topazlabs.com/topaz-video/project-starlight-series
  3. For programmatic processing, submit video to the Video API using the sl-1 model. Set technical parameters as needed and do not exceed 9 000 input frames. — https://developer.topazlabs.com/video-models/starlight/starlight-precise-1

Best practices

Superseded by this

  • 2025-05-13 — The old rule that Starlight is cloud-only no longer holds for the whole family. Starlight Mini received local rendering on Windows with NVIDIA GPUs, though this does not prove local support for original sl-1.
  • 2025-09-17 — The earlier two-model split between Project Starlight and Starlight Mini is obsolete. Starlight Sharp joined the product lineup.

Still unknown

  • The original experimental URL no longer proves the current interface or pricing tier. Current Topaz entry points lead to Astra, Topaz Video, and the API.
  • We have not verified the primary text of the 2025-02-06 X post directly. A dated Topaz Community post from that same day confirms its substance.
  • The exact date when Topaz officially renamed Project Starlight to Starlight Precise 1 remains unrecorded. Current documentation confirms continuity, but not the renaming date.

Sources

source title read
https://x.com/topazlabs/status/1887497602398073234 Topaz Labs announcement of Project Starlight 2026-09-05
https://app.topazlabs.com/experiments/starlight Project Starlight experiment 2026-09-05
https://community.topazlabs.com/t/the-process-of-improving-video-quality-with-project-starlight/87143 The process of improving video quality with Project Starlight 2026-09-05
https://community.topazlabs.com/t/project-starlight-video-ai-6-1-beta/87108?page=10 Project Starlight - Video AI 6.1 Beta 2026-09-05
https://community.topazlabs.com/t/topaz-video-ai-6-0-4/87385 Topaz Video AI 6.0.4 2026-09-05
https://community.topazlabs.com/t/topaz-video-ai-6-1-2-6-1-3/88231 Topaz Video AI 6.1.2 + 6.1.3 2026-09-05
https://community.topazlabs.com/t/topaz-video-ai-beta-6-2-0-0-b/89146 Topaz Video AI Beta 6.2.0.0.b 2026-09-05
https://community.topazlabs.com/t/video-ai-7-0-new-starlight-mini-local-ai-model/90631 Video AI 7.0 - NEW Starlight Mini (Local) AI Model 2026-09-05
https://community.topazlabs.com/t/topaz-video-1-0-0-new-studio-release/95523/1 Topaz Video 1.0.0 (New Studio Release) 2026-09-05
https://developer.topazlabs.com/video-models/starlight/starlight-precise-1 Starlight Precise 1 Developer Documentation
https://docs.topazlabs.com/topaz-video/project-starlight-series Starlight Series Topaz Video
https://www.topazlabs.com/starlight Project Starlight AI Video Enhancement

Agent brief

  • Subject: project:project-starlight, thread project-starlight-public-rollout, 2 dated events 2025-02-06 → 2025-02-11.
  • Practical note: Treat the Topaz Labs Experiments route from 2025-02-11 as a dated discovery path for Project Starlight. Verify the current experiment page before attempting use, because these links alone do not establish capabilities or access requirements.
  • Confidence: high. Dated supersedes above are the authority for what is obsolete.