Upscaling
Recover resolution and detail from any source.
4x
Maximum scale factor
~0.4s
Per frame at 1080p→4K (RTX 3060)
720p→4K
Common use case
Technical overview
Traditional upscaling (bicubic, Lanczos) interpolates between existing pixels. AI upscaling with Real-ESRGAN generates new detail by predicting what the high-resolution original would have looked like, trained on millions of image pairs.
The difference matters most on compressed or low-resolution sources where interpolation just makes soft content bigger. Real-ESRGAN recovers texture, edge definition, and fine detail that the encoder removed or the camera never captured at the original resolution.
anelo runs upscaling before depth estimation in the pipeline. This means the depth model sees the highest-quality version of each frame, producing more accurate depth maps and cleaner stereo output. A 720p source upscaled to 4K before depth estimation produces noticeably better stereo than estimating depth at 720p and upscaling after.
Use cases
- Upscale legacy 720p or 1080p content for 4K distribution
- Improve depth estimation accuracy by running depth on upscaled frames
- Prepare compressed sources (YouTube rips, old DVDs) for stereo conversion
- Deliver 4K output from sub-4K camera originals
- Recover detail lost to aggressive compression
Configuration
- Scale factorselect
Upscaling multiplier. 4x is standard (e.g. 1080p to 4K). 2x available for less aggressive enhancement.
Default: 4x
- Denoise strengthslider
Built-in denoising during upscale. Higher values remove compression artifacts but may soften fine texture.
Default: 0.5
- Model variantselect
Real-ESRGAN standard for live-action, Anime variant for cel animation and flat art styles.
Default: Real-ESRGAN 4x
Pipeline stages
AI-powered resolution enhancement using Real-ESRGAN architecture. Recovers detail and sharpness that traditional upscaling cannot, especially on compressed or low-resolution sources.
Real-ESRGAN 4x with configurable denoise strength. Anime-optimized variant available.
Available models
Output formats
Lossless per-frame output at the target resolution. Feeds into downstream pipeline stages.
Direct video output when upscaling is the only pipeline stage enabled.
Workflows
Archival upscale for re-release
Run a 720p broadcast master through 4x upscaling. Output is 4K with recovered detail suitable for modern streaming platforms. Can be combined with depth estimation for a 4K stereo 3D re-release.
Pre-depth upscale
The recommended pipeline order: upscale first, then depth estimation. The depth model sees the full-resolution frame and produces more accurate, detailed depth maps than working from the low-resolution original.
Batch quality recovery
Queue a folder of mixed-resolution clips. Each file is upscaled to the target resolution (2K or 4K). Preflight scoring flags sources with extreme compression artifacts that may not upscale cleanly.
Industries using this
Who uses this
Start with a free desktop install.
Desktop processing is free with no job limits. Pro adds cloud processing, watermark removal, and advanced 3D controls.