Upscale Tools
Topaz Image Upscale is the industry standard for faithful upscaling — it reconstructs textures, edges, and fine details in a way that looks genuinely high-resolution rather than AI-enhanced. Where Magnific V2 creatively reimagines and embellishes details, Topaz preserves the original image character with near-zero hallucination. This makes it the right choice for photography where authenticity matters: product photos, portraits, architectural photography, and any image that needs to look like it was captured at higher resolution, not generated.
Topaz pricing is based on the output megapixels, not the input. A small source image upscaled to 4K+ costs more than a medium-res image upscaled to 2K. The tiers are 12, 20, 40, and 140 credits depending on the final output size. Before upscaling, consider whether your use case actually needs 4K output — web images rarely benefit from anything above 2K, and the credit cost jumps significantly at higher output sizes.
Add a Tool node and select "Topaz Image Upscale." Connect your Image node output to the Tool input. Topaz has zero configurable parameters — no quality sliders, no sharpening controls, no scale factor selection. The model automatically analyzes the image content (faces, textures, edges, text) and applies content-aware upscaling algorithms tailored to what it detects. This zero-configuration design means the result is entirely determined by the source image quality.
The most powerful use of Topaz on Martini is as the last node in an image generation pipeline. Generate your image with FLUX.2, Midjourney, or Ideogram at standard resolution (where the model's creative strength is highest), then connect the output to Topaz for resolution enhancement. This two-step workflow produces images with both the creative quality of the generation model and the resolution of a professional camera. The alternative — generating at high resolution natively — often produces worse compositions because generative models are optimized for standard output sizes.
After upscaling, zoom to 100% and check three areas: (1) Skin and faces — Topaz reconstructs pores and fine facial detail naturally without the waxy smoothing that cheaper upscalers produce. (2) Text and labels — product labels, signage, and small text should become sharper and more readable. (3) Fabric and material textures — thread patterns, leather grain, and metal brushing should show enhanced but natural-looking detail. If any area looks artificial or over-sharpened, the source image may be too low-quality for upscaling — try regenerating it at a slightly higher base resolution.
Cost tiers are based on output megapixels: 12 credits (small), 20 credits (medium), 40 credits (large), 140 credits (very large). Check the credit cost before generating — a 256×256 thumbnail upscaled to 4K+ can cost 140 credits.
Topaz has zero parameters. If the result is unsatisfactory, the only lever is improving the source image quality — a better input always produces a better upscale.
The ideal pipeline for print-quality images: Generate with FLUX.2 or Midjourney → Upscale with Topaz. This combination produces better results than generating at 4K natively.
For portraits and headshots, Topaz is the only safe choice — Magnific V2 may alter facial features, while Topaz preserves them faithfully.
Topaz produces the most faithful upscaling of any model on Martini — it adds genuine-looking detail without hallucinating new content. The output looks like the image was originally captured at higher resolution. Compare this to Magnific V2, which actively embellishes and reimagines details (better for AI art and illustrations, but inappropriate for photography where accuracy matters). For AI-generated images, the Generate → Topaz pipeline is the recommended workflow: it combines the creative quality of the generation model with the resolution fidelity of Topaz.
Connect Topaz Image Upscale with other AI models on Martini's infinite canvas. No GPU required — start free.
Get Started Free