P-Video-Upscale.

P-Video-Upscale Overview

P-Video-Upscale is Pruna's video upscaling model in the P-Video family, available through each::labs. It increases the resolution of existing footage while preserving the source aspect ratio and audio. Its main strength is a focused, prompt-free finishing workflow: provide a video, choose a resolution target, and retrieve an enhanced MP4.

Use it for AI video enhancement, preparing low-resolution clips for larger displays, or improving footage before a final export. The four targets cover 720p, 1080p, 2k, and 4k. This is not a scene-generation or instruction-based editing tool. The source supplies the composition and timing; the model works on the footage already present. An optional Turbo setting trades some quality for faster processing.

Capabilities

• Upscale existing footage using four selectable resolution presets, including 4k for supported clips.
• Preserve the source aspect ratio for portrait, landscape, square, and other frame shapes.
• Carry source audio into the enhanced video without requiring a separate soundtrack input.
• Process live-action footage, anime, game captures, and archival clips in a video-to-video workflow.
• Offer a Turbo option for faster processing with a modest quality tradeoff.
• Accept a requested frame-rate setting from 16 to 60, with a default of 24.
• Avoid shrinking source videos that already exceed the selected resolution target.
• Return an MP4 result suitable for downloading and reviewing before delivery.

Use Cases for P-Video-Upscale

Creators: improve a short portrait video for social publishing while keeping its vertical composition and original audio. A brief such as "Prepare this creator clip at 1080p" maps to a video URL and the 1080p target, not a text prompt.

Marketers: prepare existing product footage for a higher-resolution campaign export. Compare lettering and fine packaging details against the original before approving the result.

Developers: add a finishing stage to an automated video pipeline through the P-Video-Upscale API. Select a target, validate file size and frame limits, then retrieve the completed output for downstream editing or delivery.

Designers: enhance a short game capture, anime shot, or archival insert for a larger presentation. The retained source shape helps it fit the existing layout without adding an aspect-ratio instruction.

Tips and Tricks

Start with the cleanest available source rather than a repeatedly compressed export. Keep the same input when comparing target resolutions, and inspect moving subjects, fabric, hair, and lettering at normal playback speed as well as in still frames. Upscale near the end of the edit so later resizing does not discard the resolution increase.

No text prompt is accepted. Treat requests such as "Prepare this portrait clip for a 1080p social post" or "Upscale this short landscape shot for a 4k presentation" as workflow briefs, not API parameters. Translate them into video and target. Leave turbo false for a quality-focused comparison, then try true if processing speed matters. Set fps deliberately and review the returned motion.

Technical Specifications

• Input: one video URL; maximum file size 512 MiB.
• Output: an H.264 MP4 video, delivered as a result URL.
• Resolution: 720p, 1080p, 2k, or 4k; default 1080p.
• Geometry: source aspect ratio retained; targets are pixel-area presets, not fixed dimensions for every shape.
• Frame rate: integer fps from 16 to 60; each::labs default 24.
• Length: processed-frame caps of 2,400, 1,200, 600, and 300 for the respective targets.
• Audio: source audio preserved.
• Speed: Turbo is optional; processing time varies with the request. No fixed completion time is guaranteed.

Things to Be Aware Of

Resolution labels describe pixel-area presets. A portrait or square result will not have the same width and height as a landscape result at the same target. A source larger than the selected target is not downscaled.

Clip eligibility depends on processed frames, not just file size or duration. A video under 512 MiB can still exceed its target's frame cap and be rejected. Inspect small text carefully: unreadable source lettering can become incorrect letters after enhancement. Compare the result against the original before publishing branded footage.

Key Considerations

Choose P-Video-Upscale when the shot already works and needs a resolution increase. Supply an accessible video URL, check the 512 MiB ceiling, and select a target that fits the clip's processed-frame budget. Higher targets allow fewer processed frames.

Current each::labs pricing is $0.01 per returned-video second at 720p, $0.02 at 1080p, $0.04 at 2k, and $0.08 at 4k, with duration rounded up to whole seconds. Frame rate and Turbo do not change these catalog rates. Test a short representative segment before committing to a higher-resolution delivery.

Limitations

This model requires an existing video; it cannot generate footage from a prompt or still image. There are no prompt, negative-prompt, custom-dimension, aspect-ratio, or duration inputs. It does not extend a clip, rewrite a scene, or create a soundtrack for silent footage. The highest selectable target is 4k. Inputs above 512 MiB or the selected processed-frame cap are unsupported. Enhancement cannot guarantee faithful recovery of detail missing from the source.