This project is a modified version of nekotrix project auto-boost-algorithm version 2.5.
Requirements:
- Vapoursynth
- LSMASHSource
- Av1an
- vszip
- tqdm
- psutil
- mkvmerge
Optionally:
- turbo-metrics
- vship
- ffmpeg with XPSNR
- Added a progressbar to the turbo-metrics calculation
- Added a stage 4 for a final encode with Av1an
- Removed the ffmpeg dependency for XPSNR calculation
- Adjustable aggressiveness
- Better metrics implementations options
- Temporarly removed custom skip
- (Technically more accurate by a small margin)
- (WIP) Add a home-made encoding framework (alternative to Av1an), because why not
- Improve CLI
- (Maybe) Add a TUI
- (Some day) Add a GUI
- Make/Keep the code fully pep8 compliant
- Refactor the code to have a Model-View-Controller architecture
| Parameter | Default | Possible values | Explanation |
|---|---|---|---|
-i, --input |
Video input file path | ||
-t, --temp |
Input filename | Temporary directory | |
-s, --stage |
0 |
0, 1, 2, 3, 4 |
Boosting stage (All, fast pass, metrics calculation, zones generation, final encode) |
-crf |
30 |
1-63 |
CRF to use |
-d, -deviation |
10 |
0-63 |
Maximum crf deviation from base crf |
--max-positive-dev |
0-63 |
Maximum positive deviation from base crf | |
--max-negative-dev |
0-63 |
Maximum negative deviation from base crf | |
-p, --preset |
6 |
-1-13 |
SVT-AV1 preset to use |
-w, --workers |
CPU core count | 1-Any |
Number of Av1an workers |
-m, --method |
1 |
1, 2, 3, 4 |
Zones calculation method (1: SSIMU2, 2: XPSNR, 3: Multiplied, 4: Minimum) |
-a, --aggressiveness |
20 or 40 |
0 - Any |
Choose boosting aggressiveness |
-M, --metrics-implementations |
vship,vszip |
<vszip, vship, turbo-metrics>,<vszip, ffmpeg> |
Choose each metric implementation. First is SSIMULACRA2, second is XPSNR |
-v --video-params |
Encoder parameters for Av1an | ||
-ef, --encoder-framework |
av1an |
av1an |
Encoding framework to use |
-o, --output |
Input file directory | Output file for final encode |
This project is based on the original work by nekotrix.
Original contributors include R1chterScale, Yiss, Kosaka, and others.
Licensed under MIT, see LICENSE.