Upscale video with Topaz's Astra 2 creative model. Enlarge up to 4x and 4K with new generated detail. Try free — pay per generation, no subscription. Upscale a video with Topaz's Astra 2 creative diffusion model, generating new plausible detail while enlarging up to 4x and up to 4K.
Upscale a video with Topaz's Astra 2 creative diffusion model, generating new plausible detail while enlarging up to 4x and up to 4K.
About this model
Topaz Upscale Video Creative runs Topaz's Astra 2 diffusion model to enlarge video up to 4x — including up to 4K output — while generating new, temporally consistent fine detail rather than just interpolating pixels. Creativity, realism, and sharpness are all independently adjustable, and an optional text prompt can guide what kind of detail gets added (note: setting a prompt limits the input to 450 frames). Frame rate can be raised up to 60fps via target_fps. For a more literal, detail-preserving upscale of clean sources, see Topaz Upscale Video Precision; for degraded or low-resolution footage that needs realistic reconstruction, see Topaz Upscale Video Generative.
Cost analysis
| Provider | Cost | Notes |
|---|---|---|
| muapiapp | $3.00-$5.00+ per 10 seconds of output depending on resolution (plus margin), billed dynamically | Pay-per-use, no subscription. Cost is calculated from the real output resolution/duration at request time. |
| Fal.ai | Not available | Not offered as a standalone public API by muapiapp's upstream infrastructure partner. |
| Replicate | Not available | Not currently listed on Replicate. |
Pay-per-use, no subscription. Cost is calculated from the real output resolution/duration at request time.
Not offered as a standalone public API by muapiapp's upstream infrastructure partner.
Not currently listed on Replicate.
** Competitor pricing is estimated based on similar model architectures and usage tiers.
Configuration schema
| Parameter | Type | Description | Default |
|---|---|---|---|
| Video URL | string | Input video to upscale. | https://d3adwkbyhxyrtq.cloudfront.net/webassets/videomodels/topaz-upscaler-in.mp4 |
| Prompt | string | Optional text prompt to guide creative detail (limits input to 450 frames when set). | - |
| Creativity | number | Creative detail strength (0-1). | 0.5 |
| Realism | number | Realism strength (0-1). | - |
| Sharp | number | Sharpening amount (0-1). | - |
| Upscale Factor | number | Factor to upscale the video by. | 2 |
| Target FPS | int | Target output frame rate. | - |
| H264 Output | boolean | Encode output using H.264 instead of the default codec. | false |
Input video to upscale.
https://d3adwkbyhxyrtq.cloudfront.net/webassets/videomodels/topaz-upscaler-in.mp4Optional text prompt to guide creative detail (limits input to 450 frames when set).
-Creative detail strength (0-1).
0.5Realism strength (0-1).
-Sharpening amount (0-1).
-Factor to upscale the video by.
2Target output frame rate.
-Encode output using H.264 instead of the default codec.
falseDeveloper documentation
Provide a video: Set video_url to a publicly accessible video file URL.
Set the scale: upscale_factor accepts 1-4x; the resulting resolution tier (up to 1080p vs. 4K) drives pricing.
Tune creative controls: creativity (0-1, default 0.5), realism, and sharp shape how much new detail is generated.
Optionally guide it: Add a prompt to steer the added detail (limits input to 450 frames), and set target_fps (16-60) to raise frame rate.
Submit and poll: POST to /api/v1/topaz-upscale-video-creative, then poll /api/v1/predictions/{request_id} or use webhook_url — video jobs can take several minutes.
curl -X POST https://api.muapi.ai/api/v1/topaz-upscale-video-creative \
-H "x-api-key: YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"video_url": "https://example.com/clip.mp4",
"upscale_factor": 2,
"creativity": 0.5
}'
Frequently asked
It enlarges a video up to 4x (including up to 4K) using Topaz's Astra 2 diffusion model, generating new temporally consistent detail rather than simple pixel interpolation.
Creative (Astra 2) leans furthest toward generating new stylized detail. Generative (Starlight models) reconstructs realistic detail for degraded footage, and Precision (Proteus/Artemis/etc.) stays closest to the source.
Yes, set `target_fps` (16-60) to output at a higher frame rate than the source; higher fps increases cost proportionally.
Setting `prompt` guides the added creative detail, but limits the processed input to 450 frames.
Cost is billed per 10 seconds of output at a 30fps baseline (higher `target_fps` scales cost up proportionally), with a higher rate for 4K output than for up to 1080p.
Yes, set `H264_output: true` to encode the result as H.264 instead of the default codec.