Preprocessing
Clean, correct, and prepare before the heavy lifting.
5
Preprocessing operations
<0.05s
Per frame (non-AI stages)
First
Runs before all AI stages
Technical overview
Every downstream AI model is sensitive to input quality. Noise in the source becomes micro-structures in the depth map, which become distracting shimmer in stereo output. Compression artifacts become false edges that the depth model treats as real geometry. Color inconsistency between frames causes the stereo pair to mismatch.
anelo applies preprocessing in an opinionated order designed to maximize downstream quality: denoise first (removes noise before sharpening amplifies it), then deartifact (repairs compression damage), then sharpen (enhances real detail, not noise), then color correction and LUT application (ensures consistent color before stereo pairing).
Grain management is intentionally separate from denoising. Film grain is an aesthetic choice — some content needs it preserved or enhanced, other content needs it stripped. anelo lets you remove grain before processing (cleaner depth maps) and optionally re-add matched grain after stereo compositing (preserved aesthetic).
Use cases
- Clean noisy footage before depth estimation to prevent false depth structures
- Remove compression artifacts from streaming or compressed sources
- Apply color correction and LUTs for consistent stereo output
- Sharpen soft sources without amplifying noise
- Strip and re-apply film grain for aesthetic control through the pipeline
Configuration
- Denoise methodselect
hqdn3d for speed, nlmeans for quality. nlmeans preserves more detail but is significantly slower.
Default: hqdn3d
- Denoise strengthslider
Aggressiveness of noise removal. Higher values remove more noise but may soften fine detail.
Default: 0.5
- Sharpen methodselect
CAS (contrast-adaptive) for general use, unsharp mask for aggressive detail recovery.
Default: CAS
- Deartifacttoggle
Removes compression blocking and ringing artifacts. Most useful on heavily compressed sources.
Default: Off
- LUT fileselect
Apply a .cube LUT for color correction. Applied after denoise and sharpen, before stereo stages.
Default: None
Pipeline stages
Detects shot boundaries, motion complexity, and optimal processing parameters before any frames are touched. This preflight pass prevents wasted compute and flags potential issues early.
Adaptive sensitivity with configurable shot boundary detection.
Extracts individual frames as lossless images for per-frame AI processing. Preserves the full color space and bit depth of the source material.
PNG lossless output in sRGB. Supports 8/10/12-bit sources.
Output formats
Cleaned frames ready for downstream pipeline stages.
Direct video output when preprocessing is the only pipeline stage enabled.
Workflows
Source cleanup before stereo conversion
Run denoise and deartifact on compressed sources before the depth estimation stage. This prevents noise and blocking from creating false depth structures that appear as shimmer or geometry errors in the stereo output.
Color-correct for stereo comfort
Apply a LUT or manual color correction before stereo warping. Both eyes receive identical grading, preventing the color mismatch that causes visual fatigue in 3D viewing.
Grain round-trip
Remove grain before processing, run the full pipeline on clean frames, then re-add matched grain to the final output. The depth model sees clean data, but the output preserves the original aesthetic.
Industries using 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.