Scenario provides a range of image upscaling models that enhance images by increasing their resolution while preserving important details. Each model is tuned for a particular style of upscaling—some prioritise fidelity and realism while others add creative or stylised elements. Use these references to select the right model and understand its typical scaling range. When calling the /generate/custom/{modelId} endpoint, supply the appropriate modelId and parameters for your chosen model. Details on each model’s supported parameters are provided below.

Scenario Upscalers

Scenario Flux Upscale

Model ID: model_sc-upscale-flux

Model purpose: A versatile, balanced model designed for high‑quality upscaling. It supports LoRA models for consistent styles or characters and accepts style images. You can provide natural language prompts and advanced parameters for fine‑tuned results.

• Typical scaling factors: 2×–8×.
• Recommended use: When you need flexible control over fidelity and style, such as upscaling concept art or illustrations.

Scenario Upscale V3

Model ID: model_upscale-v3

Model purpose: A speed‑optimised model based on SDXL that offers efficient upscaling while maintaining good quality.

• Typical scaling factors: 1×–16×
• Recommended use: Quick upscaling when cost efficiency and speed are important.

Scenario Texture Upscale

Model ID: model_sc-upscale-flux-texture

Model purpose: Enhances surface details and material textures such as fabrics, metals or terrain.

• Typical scaling factors: 2×–8×.
• Recommended use: When refining materials in environmental assets or game textures.

Scenario Skybox Upscale

Model ID: model_sc-upscale-flux-skybox

Model purpose: Tailored for panoramic or 360° backgrounds; it maintains scene consistency across wide fields of view when resolution increases.

• Typical scaling factors: 2×–8×.
• Recommended use: Upscaling skyboxes and panoramic environment textures.

---

Topaz

Topaz Image Upscale

Model ID: model_topaz-image-upscale

Model purpose: A professional‑grade upscaler from Topaz Labs that focuses on fidelity and realism. It offers several internal modes (Standard V2, Low Resolution V2, CGI, High Fidelity V2, Text Refine) tailored to different input types.

• Typical scaling factors: 2×–6×
• Recommended use: When you need high‑fidelity results, such as enhancing photo textures or production assets.

---

Recraft

Recraft Crisp Upscale

Model ID: model_recraft-crisp-upscale

Model purpose: A fast, budget‑friendly upscaler that provides instant resolution enhancement. It emphasises sharpness and clean edges with no extra parameters to set.

• Typical scaling factor: 4× fixed.
• Recommended use: Rapid improvement of small assets or icons where crisp edges are important.

Recraft Creative Upscale

Model ID: model_recraft-creative-upscale

Model purpose: Automatically adds unique artistic touches while increasing. It’s essentially a creative counterpart to Recraft Crisp.

• Typical scaling factor: 4× fixed.
• Recommended use: Stylised assets, concept art or where creative interpretation is welcome.

---

Bytedance

Seed VR2 Upscale Image

Model ID: model_seedvr-2-upscale-image

Model purpose: An experimental model based on the Seed VR2 diffusion architecture that increases resolution while preserving texture fidelity and minimising artifacts. It uses a one‑step diffusion transformer similar to the video variant and can recover details in degraded images.

• Typical scaling factors: 2×–10×.
• Recommended use: Restoring blurred or low‑detail images, or when large upscales are needed.

---

Clarity AI

Crystal Upscaler

Model ID: model_crystal-upscaler

Model purpose: A high‑precision AI model built to enlarge visual assets with minimal quality loss. Suitable for both subtle and extreme upscaling.

• Typical scaling factors: 1×–16×
• Recommended use: General‑purpose upscaling across a wide range of subjects.

---

Note: The above parameters are typical for image upscaling models. Some models may expose additional parameters or omit certain options. Refer to each model’s API documentation (or contact Scenario support) for full details.

---