Hello 
Back with another API release thread for the month of April!
This release is model-focused, adding six new models to the API (four image models and two video models), expanding what’s available across various model families
New Models
Six new models are now available in the API:
Wonder 3 — Most advanced generative realism model for upscaling images
As a significant improvement from Wonder 2, this new model has proven to perform better in a few key areas—namely, providing more natural detail and improved realism from its previous version, which yielded many edge cases with plasticky skin texture and artifacts. Wonder 3 performs across all resolution levels and is well suited for portraits, landscapes, wildlife, and complex textures such as fabrics, metallics, feathers, and skin.
Denoise Max — Highest-quality denoising with intelligent detail recovery
This is our first generative denoise model, using some of the same advancements from the Wonder series to intelligently sharpen blurry areas while leaving already-sharp areas untouched. It yields significantly less artifacts in comparison to our previous Denoise models.
High Fidelity 3 — Improved fidelity upscaling for high-resolution images and professional photos
This model improves on High Fidelity 2 by delivering enhanced resolution scaling and minimal detail loss or artifacting. It focuses on preserving original image structure and realism, reducing the plasticky appearance and excess artifacts seen in earlier versions while maintaining a clean, natural output. This model performs well on a wide range of content such as portraits, landscapes, wildlife, and images with graphics.
Super Focus 3 — Generative blur recovery with selective refinement
Super Focus 3 is a next-generation model designed to recover detail from severely blurred or out-of-focus images when traditional sharpening methods fall short. This is the newest version of our Super Focus (generative sharpening) model line, particularly improving on Super Focus 2 by delivering improved fidelity, especially on faces and skin.
Astra 2 — Enhanced creative upscaling for GenAI video with advanced control over detail and style
Astra 2 builds on Astra 1 (slc-1), our most creative model, with expanded control over detail generation and stylization. It transforms AI-generated footage by adding new, dynamic detail and texture, with adjustable creativity and sharpness controls and prompt-based guidance for more precise creative direction. This model performs incredibly well on stylized and GenAI sources, as well as wide shots and large-scale scenes where additional detail can be creatively inferred.
Hyperion (SDR to HDR) — Unlock HDR depth from SDR footage
Hyperion converts standard dynamic range (SDR) video into high dynamic range (HDR) by expanding luminance range, bit depth, and contrast. It redistributes tonal values across highlights, midtones, and shadows to better utilize HDR display capabilities while preserving image detail and balance. Note that before using this model, we highly recommend checking out the docs page for it, as we’ve added a section with details on supported codecs, important considerations, and more—check out those sections here!
We recommend referencing the linked model pages above to view full documentation that includes model descriptions, pricing, endpoints / parameters, and output examples.
As always, we’d love to hear your feedback, questions, and ideas! If you’d like to get involved, feel free to leave comments here in our community forum or in our Developer Discord!
Changelog
-
Launched six new models (four image, two video)
-
Wonder 3
-
Denoise Max
-
High Fidelity 3
-
Super Focus 3
-
Astra 2
-
Hyperion (SDR to HDR)
-
For any issues, reach out to support@topazlabs.com. Enterprise inquiries can be directed to enterprise@topazlabs.com.
Aiden Warnke
Product Manager
Enterprise & API
