A video upscale normally finishes within 1 to 5 minutes. If it's stuck longer than 5 minutes, hover over the job tile in Assets and select Cancel, then retry. The most common causes of failure are an unsupported input format, source video quality too low to reconstruct detail, or queue overload. Credits for failed upscales are returned automatically: verify in Manage Account → Usage.
What are normal upscale wait times?
A video upscale typically completes within 1 to 5 minutes depending on source video length, resolution, and the chosen model. If it's stuck for more than 5 minutes, cancel and retry.
Which upscale model should I choose?
To upscale, go to https://higgsfield.ai/ → Video → Upscale, or open any video in Assets and select Upscale. The same tool handles images.
For videos, two models are available:
Model | Best for |
|---|---|
ByteDance Upscale | High-fidelity upscaling up to 8K with preset-tuned detail |
Topaz Video | A diffusion-based model that adds creative details |
For images, three models are available:
Model | Best for |
|---|---|
Topaz | The default for general-purpose upscaling |
ByteDance Upscale | High-resolution upscaling up to 4K |
Topaz Generative | A diffusion-based model that adds creative details on its own |
The rule of thumb is the same for both: ByteDance and the default Topaz sharpen and enhance what's already there, while the diffusion-based options (Topaz Video for videos, Topaz Generative for images) can invent fine detail while upscaling. Use the diffusion models for creative work, and avoid them when the output must stay exactly as generated: product shots, text-heavy visuals, or client deliverables that were already approved.
Set the target quality with the Scale factor selector: the available options depend on the selected model, up to 8K with ByteDance Upscale for videos. Additional options are available under Advanced settings, and the credit cost updates on the Upscale button as you change them.
What are the most common causes of upscale failure?
Unsupported input format — use a standard format (MP4 with H.264 encoding) for best compatibility.
Source video quality too low — extremely low resolution or heavily compressed source videos may not upscale successfully: the model needs sufficient visual information to reconstruct detail.
Queue overload — during high platform load, upscale jobs may queue longer than usual.
How do I cancel a stuck upscale?
Hover over the upscale job tile in Assets: the Cancel button appears in the top-right corner. Select it, then retry.
Will I get my credits back for a failed upscale?
Yes. Credits for failed upscale jobs are returned automatically. Verify in Manage Account → Usage. If credits were not returned, contact us through higgsfield.ai/contact with the generation ID: you can copy it from the generation's page URL in Assets.
What if the issue persists?
Check that your source video is in a standard format (MP4, H.264), and try a different upscale model. For browser-level issues, see What should I do if the Higgsfield site, Canvas, or Supercomputer won't load?