Pruna P-Video-Upscale Video Upscaler

Overview

P-Video-Upscale is a video-to-video upscaling model by Pruna. You upload one source clip, pick a target resolution preset, and the model increases resolution while keeping the original aspect ratio. It returns a new H.264 MP4 and keeps the source audio track, so you can drop the result back into an edit timeline without re-sync work. It can also interpolate frames when you request a higher frame rate than the input, which helps when you need smoother motion delivery.

What you can build

• Upscale short marketing clips for 1080p or 4K delivery
• Improve clarity on AI-generated video before publishing
• Prep footage for post-production workflows that require a minimum resolution
• Create sharper product demos from older low-resolution exports
• Upscale archived social content while preserving the original audio

Inputs

• A single video file to upscale, up to 512 MiB. The model keeps the clip’s aspect ratio and keeps its audio track.
• A target resolution preset: 720p, 1080p, 2K, or 4K. If the source video is already larger than the chosen target, it won’t be reduced.
• An optional output frame rate from 16 to 60 fps. Leave it unset to keep the input video’s frame rate (up to 60 fps). If you set a higher rate than the source, the model interpolates frames.
• An optional fast mode that runs with slightly lower visual quality.
• An optional safety check that screens the clip against the content policy before upscaling.

Outputs

• One upscaled video file encoded as an H.264 MP4.
• The output keeps the source aspect ratio.
• The output keeps the source audio track.
• The output uses your requested frame rate, or keeps the original if you leave it unset.

Recommended settings

• For most web video: upscale to 1080p and keep the original frame rate.
• For the cleanest results: leave fast mode off.
• For smooth motion delivery: only raise FPS when you accept interpolation artifacts on fast movement.
• For longer clips: choose a lower target resolution, or reduce FPS, to stay under the frame cap.

Limitations

• Input size is limited to 512 MiB per clip.
• Output frame rate is limited to 16–60 fps.
• Maximum clip length depends on how many frames the run processes. At 30 fps or higher, typical limits are about 80s at 720p, 40s at 1080p, 20s at 2K, and 10s at 4K.
• The model won’t downscale a source that exceeds your chosen target.
• If you increase FPS above the source rate, the model interpolates frames. This can cause motion artifacts on fast cuts, particles, or thin lines.
• Low-quality inputs reduce results. Heavy compression, strong noise, banding, interlacing, or hard-burned subtitles can limit how much detail the upscale can recover.

Safety & compliance

• You can enable a safety check that screens the clip against the content policy before processing.
• Some categories are blocked regardless of the safety toggle.
• You’re responsible for having rights to the video and audio you upload, and for how you use the output.
• Don’t use the model to create or enhance disallowed content, including illegal sexual content or self-harm instructions.