Upscale images with Topaz Wonder & Recover models. Reconstruct realistic detail, textures, and faces at high resolution. Try free — pay per generation.
Upscale images with Topaz Wonder & Recover models. Reconstruct realistic detail, textures, and faces at high resolution. Try free — pay per generation.
关于此模型
Topaz Upscale Image Generative reconstructs realistic fine detail — skin texture, fabric, foliage, faces — using Topaz's Wonder and Recover model families. Unlike a purely creative upscaler, Wonder models aim to recover what a higher-resolution capture would plausibly have looked like, with dedicated face-enhancement controls and subject-detection to focus enhancement on foreground, background, or the whole frame. Nine model variants (Wonder 3.5/3/2, Recover 3, Standard MAX, Redefine, Recovery V2/Recovery) cover everything from general photo restoration to heavier reconstruction of badly degraded sources. Pairs well with Topaz Upscale Image Precision for a lighter-touch faithful pass, or Topaz Upscale Image Creative for a more stylized result.
成本分析
| 提供商 | 费用 | 备注 |
|---|---|---|
| muapiapp | $0.08+ per started 4-8 megapixels of output depending on model (plus margin), billed dynamically | 按用量付费，无订阅。费用根据请求时的实际输出分辨率/时长计算。 |
| Fal.ai | 暂不可用 | muapiapp 的上游基础设施合作方目前未将其作为独立的公共 API 提供。 |
| Replicate | 暂不可用 | 目前未在 Replicate 上架。 |
按用量付费，无订阅。费用根据请求时的实际输出分辨率/时长计算。
muapiapp 的上游基础设施合作方目前未将其作为独立的公共 API 提供。
目前未在 Replicate 上架。
** 竞品价格根据相似模型架构和使用层级估算。
配置参数
| 参数 | 类型 | 描述 | 默认值 |
|---|---|---|---|
| 图像 URL | string | Input image to upscale. | https://d3adwkbyhxyrtq.cloudfront.net/webassets/imagemodels/topaz-upscaler-in.jpg |
| 模型 | 枚举（9 个选项） | Topaz generative upscale model variant. | Wonder 3 |
| 放大倍数 | number | Factor to upscale the image by. | 2 |
| Crop To Fill | boolean | Crop the output to exactly fill the target aspect ratio. | false |
| 输出格式 | 枚举（2 个选项） | Output image file format. | jpeg |
| Subject Detection | 枚举（3 个选项） | Which region of the image to prioritize enhancement on. | All |
| Face Enhancement | boolean | Enable dedicated face enhancement. | true |
| Face Enhancement Strength | number | Strength of face enhancement. | 0.8 |
| Face Enhancement Creativity | number | Creativity applied to face enhancement. | 0 |
| Enhancement Strength | 枚举（3 个选项） | Overall enhancement strength tier. | - |
| 创意度 | int | Creative detail strength (1-6). | - |
| 纹理 | int | Texture detail strength (1-5). | - |
| 提示词 | string | 可选的 text prompt to guide generative detail. | - |
| Autoprompt | boolean | Automatically generate a guiding prompt. | - |
| Sharpen | number | Sharpening amount (0-1). | - |
| Denoise | number | Denoising amount (0-1). | - |
| Detail | number | Fine-detail amount (0-1). | - |
Input image to upscale.
https://d3adwkbyhxyrtq.cloudfront.net/webassets/imagemodels/topaz-upscaler-in.jpgTopaz generative upscale model variant.
Wonder 3Factor to upscale the image by.
2Crop the output to exactly fill the target aspect ratio.
falseOutput image file format.
jpegWhich region of the image to prioritize enhancement on.
AllEnable dedicated face enhancement.
trueStrength of face enhancement.
0.8Creativity applied to face enhancement.
0Overall enhancement strength tier.
-Creative detail strength (1-6).
-Texture detail strength (1-5).
-可选的 text prompt to guide generative detail.
-Automatically generate a guiding prompt.
-Sharpening amount (0-1).
-Denoising amount (0-1).
-Fine-detail amount (0-1).
-开发者文档
Provide an image: Set image_url to a publicly accessible image URL.
Choose a model: Wonder 3 is the default general-purpose choice; use Wonder 3.5 for the newest model, or Recover 3/Recovery V2 for heavily degraded sources.
Tune face and detail controls: face_enhancement, face_enhancement_strength, creativity, texture, sharpen, and denoise all shape the reconstruction.
Optionally guide it: Add a prompt (or set autoprompt: true) to steer what detail gets reconstructed.
Submit and poll: POST to /api/v1/topaz-upscale-image-generative, then poll /api/v1/predictions/{request_id} or use webhook_url.
curl -X POST https://api.muapi.ai/api/v1/topaz-upscale-image-generative \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"image_url": "https://example.com/portrait.jpg",
"model": "Wonder 3",
"upscale_factor": 2,
"face_enhancement": true
}'
常见问答
It reconstructs realistic fine detail (textures, faces, foliage) while upscaling an image up to 4x, using Topaz's Wonder and Recover model families.
Wonder 3 (default) is a solid general choice; Wonder 3.5 is the newest revision; Recover 3 and Recovery V2 target heavily degraded or low-quality sources.
Yes, `face_enhancement` defaults to true with `face_enhancement_strength` at 0.8 — both are adjustable, and `face_enhancement_creativity` controls how much new facial detail is invented.
Yes — `subject_detection` accepts `All`, `Foreground`, or `Background` to focus enhancement on a specific region.
Wonder 3/3.5 bill per started 8-megapixel block of output; all other models bill per started 4-megapixel block, both based on the real output size after `upscale_factor`.
Yes, the optional `prompt` field used to steer reconstruction accepts up to 1024 characters.