Video Upscaling
AI-powered video upscaling with Real-ESRGAN and SwinIR models. Enhance resolution, reduce noise, and sharpen details with a professional-grade processing pipeline.
Recaster includes a complete video upscaling system that uses state-of-the-art AI models to increase video resolution while preserving detail. Whether you need a quick 2x upscale for a preview or a maximum-quality 8x enlargement for final delivery, the upscaling tools give you full control over every stage of the process.
Upscaling Models
Compare Real-ESRGAN and SwinIR models. Understand quality ratings, file sizes, tier availability, and platform compatibility.
Quality Presets
Six optimized presets for different use cases: Fast, Balanced, Quality, Anime, Restoration, and Temporal.
Pipeline Settings
Configure the multi-stage processing pipeline: denoising, multi-pass upscaling, color correction, sharpening, and tile settings.
Remote Upscaling
Upscale videos on cloud GPUs with the Studio tier. Upload, process, and download results with full pipeline support.
How It Works
The upscaling system uses a multi-stage pipeline that processes each video frame through several configurable stages:
- Pre-Processing — Optional denoising removes grain and noise before upscaling, which prevents the AI model from amplifying artifacts.
- Upscaling — The core AI model (Real-ESRGAN or SwinIR) increases the resolution. Multi-pass mode chains multiple 2x passes for better quality at higher scales.
- Post-Processing — Optional color correction preserves the original color palette, and sharpening adds edge definition to the upscaled output.
- Finalization — The original audio track is automatically muxed back into the upscaled video, and metadata such as FPS and duration is preserved.
Drag-and-Drop Interface
To upscale a video, click the Upscale button in the sidebar, then drag and drop your video file onto the drop zone. You can also click the drop zone to browse for a file. The interface displays the input resolution and automatically calculates the output resolution based on your chosen scale factor.
Supported Formats
The upscaling system accepts common video formats including MP4, AVI, MOV, MKV, and WebM. Output is written as MP4 with the original audio track preserved. For best results, use source material with as little compression as possible.
Audio Preservation
Available Scale Factors
Three scale factors are available depending on the model and your license tier:
| Scale | Example | Tier |
|---|---|---|
| 2x | 1920x1080 → 3840x2160 | Free |
| 4x | 1920x1080 → 7680x4320 | Free |
| 8x | 1920x1080 → 15360x8640 | Studio |
Performance
Upscaling speed depends on your GPU, the chosen model, and the input resolution. As a general guide, expect approximately 5 to 10 frames per second on a modern GPU such as an RTX 3090 or RTX 4090. Small images under 512 pixels are processed directly, while larger frames are automatically split into tiles for efficient memory usage.
GPU memory usage is typically around 2 GB for 4x upscaling. The tile size, overlap, and GPU memory limit can be adjusted in the advanced settings panel to match your hardware.
Next Steps
Start by reviewing the available upscaling models to choose the best one for your project. Then explore the quality presets for quick, optimized configurations, or dive into the pipeline settings for full manual control.
Was this page helpful?