Topaz Image Upscale
What is Topaz Image Upscale?
Topaz Image Upscale is a single-call image enhancement and resizing endpoint backed by Topaz Labs' production model. The API mirrors the upscaling pass inside Photo AI: pick a magnification factor from 1x through 16x, optionally adjust denoise and sharpen between 0.0 and 1.0, and toggle face enhancement when the source is a portrait. There is no separate fast-or-quality decision to make. Topaz applies its production engine on every call, and the controls you pass select what the pass does, not which model runs it. Output is delivered as a downloadable image URL once the task completes.
Key features of Topaz Image Upscale
Five features cover the surface area you'll actually use.
Five magnification levels, including 1x
Pick 2x, 4x, 8x, or 16x for real magnification, scaling the longest sides linearly. Pick 1x to keep output dimensions identical to input and run only the enhancement pass. Useful when the resolution is fine but the source is noisy or soft.
Denoise and sharpen as separate sliders
Both run from 0.0 to 1.0, defaulting to 0.5. Treat them as independent dials. Compressed JPEGs reward low denoise and moderate sharpen; clean digital sources can take more sharpen without halo. The split lets you tune for material instead of forcing a single quality knob.
Face enhancement tuned for portraits
On by default, the face pass recovers detail in eyes, hair, and skin texture without smoothing the rest of the frame. Turn it off for landscapes, product shots, and non-human subjects, where it can amplify random textures into face-like artifacts.
Material-agnostic input handling
Photographs, screenshots, anime art, line work, document scans, and game assets all flow through the same endpoint. Adjust the sliders rather than the model: sharpen high for line work and lettering, low for photographs, and 1x for in-place restoration without resizing.
16x ceiling for print and large displays
A 1024-pixel-wide source becomes 16384 pixels at 16x. Enough headroom for billboard prints, large-format event displays, and high-DPI master files. Plan downstream storage around the resulting size and pick a smaller factor when delivery requirements are kinder.
Best for
E-commerce catalogs
Lift small supplier JPEGs to storefront-ready sizes with 4x or 8x upscaling, denoise pulled down to keep texture, and face enhancement off for product-only shots.
Print prep from digital-only sources
4x or 8x passes when the design file is all that exists. Push sharpen toward 0.7 for line work and lettering, ease back for photographs.
Portrait restoration and headshot prep
Face enhancement on by default. Pair with 2x or 4x and a moderate denoise value to lift older portraits before manual retouching.
Document and screenshot scaling
2x with sharpen tuned to ~0.6 cleans up screenshots, scanned forms, and OCR-bound pages without softening text edges.
Anime, CG, and game asset scaling
8x or 16x with sharpen turned up keeps line clarity and color blocks intact when scaling pixel art or marketing key art.
In-place enhancement at original size
Set the upscale factor to 1x to run denoise, sharpen, and face passes without changing geometry. Useful when the resolution is already correct but the image is noisy or soft.
Use cases
Wire a supplier-image normalisation step into a product ingestion pipeline: every new SKU image flows through Topaz at 4x and lands at storefront-ready resolution before it reaches the catalog. Build a portrait restoration tool that lets users upload a phone photo of an old print and receive a face-enhanced 4x version back. Scale digital marketing key art to print resolution by sending the latest hero asset through 8x with sharpen tuned high. Add an in-place clean-up step to a content management workflow by running every uploaded image through 1x for grain reduction without any size change.
API examples
Call Topaz Image Upscale from any language by POSTing to /v1/tasks. Full parameter docs live at docs.unifically.com/models/image/topaz-labs/image-upscale.
curl -X POST https://api.unifically.com/v1/tasks \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "topaz-labs/image-upscale",
"input": {
"image_url": "https://example.com/photo.jpg",
"upscale_factor": "4x"
}
}'
Successful submission returns a task_id. Poll GET /v1/tasks/<task_id> or set a callback_url on the request to receive the finished image URL.
FAQs
People also ask
Topaz Image Upscale is the API form of the Topaz Labs upscaling engine that powers Photo AI and Gigapixel desktop. Send any image URL, choose an upscale factor from 1x through 16x, optionally tune denoise and sharpen, and the model returns a higher-resolution version with optional face enhancement applied.
Five. 1x runs enhancement only without changing pixel dimensions. 2x, 4x, 8x, and 16x scale the longest sides linearly (a 500x500 input becomes 1000x1000 at 2x and 8000x8000 at 16x). The default is 1x, so you opt into a real magnification level.
When the image is already at the resolution you need but suffers from grain, JPEG compression, mild blur, or loss of fine detail. 1x runs Topaz's denoise, sharpen, and face enhancement passes without resizing the canvas, so output dimensions stay identical to input.
Both accept a float between 0.0 and 1.0 with a default of 0.5. Higher denoise removes more grain at the cost of fine texture; lower preserves more grain. Higher sharpen recovers more edge detail but can produce halos on already-sharp sources. For compressed JPEGs, try denoise 0.3 with sharpen 0.5.
When face enhancement is on (the default), Topaz applies an extra pass that recovers detail in human faces (eyes, hair, skin texture). It is tuned for portraits and headshots. Turn it off on landscapes, product shots, or non-human subjects, where it can amplify random textures into face-like artifacts.
The output of a 16x pass on even a small input is large. A 1024-pixel-wide source becomes 16384 pixels wide. Pick a smaller upscale factor when an input combined with the chosen factor would overshoot what your delivery pipeline or downstream tools accept.
4x with denoise around 0.3 and sharpen around 0.5 is a strong starting point. It lifts small supplier JPEGs to storefront-ready sizes without the over-sharpened halo that a pure 8x pass tends to produce on already-compressed sources.
