Description 📝

Placebo for audiophiles, the successor of soundless and one of a kind software.

Features ✨

• Estimation of:
  • Sample rate
    • No AI usage by default
  • Bit depth
  • Channel count
  • Bitrate (calculation)
  • dBFS peak
• Writing to output with:
  • Cross-extension metadata copying
  • Post-processing effects:
    • Spectral gate
    • Silence removal (trim, dBFS/empty samples)

Note

For more explanations, please consider visiting frequently asked questions wiki page.

Genesis 🏛️

April 21, 1997, can be considered as the worst day in the history of music quality. The release of Winamp and the rapid rise of the MP3 format that came from it began a revolution in accessibility, but it came at a cost: music quality dropped. Lossless releases were encoded to questionable quality by untrained people. Fullband sound went to further plan, and when it resurfaced, it often came with degraded dynamics and equivalent quality. Sometimes worse.

Since 2018 a flood of so-called “sound engineers” trained by YouTube, unaware of even the basics of audio science, has filled — or rather attempted to redefine — the industry with poor recordings and narrowband samples. The true craft of music production has been buried under noise.

This is why I built axiom — not to “improve” the sound¹, but to expose what it might be.

What is the approximate frequency cutoff?

How many channels are actually there?

Is this high‑res file really what it claims to be?

¹ - although it MAY upsample fake-lossless files a bit...

Roadmap 🏁

• Bug fixing 🐛
• Writing to multiple files 🔢

Installation 🖥️

1. pip (reccomended)

python -m pip install git+https://github.com/kubinka0505/axiom -U

Usage 📝

Module (simple)

>>> from axiom import Axiom
>>> 
>>> # Create object
>>> axiom_obj = Axiom("HD.flac")
>>> 
>>> # Estimate sample rate
>>> axiom_obj.sample_rate()
{'HD.flac': {'samplerate': 33462, 'cutoff': 16731.298828125}}
>>> 
>>> # Estimate bit depth
>>> axiom_obj.bit_depth()
{'HD.flac': {'bit depth': 10, 'bit depth snapped': 16}
>>> 
>>> # Estimate channels
>>> axiom_obj.channels()
{'HD.flac': {'channels': 2}
>>> 
>>> # Estimate peak
>>> axiom_obj.peak()
{'HD.flac': {'peak': 0.0}
>>> 
>>> # All in one
>>> axiom_obj.aio()
{'HD.flac': {'samplerate': 33462, 'cutoff': 16731.298828125, 'bit depth': 10, 'bit depth snapped': 16, 'channels': 2, 'bitrate': 1070803.125, 'peak': 0.0}}

Module (advanced)

>>> from axiom import Axiom
>>> 
>>> # Files
>>> to_process = [
...     "%UserProfile%/file.mp3",
...     "~/audio.wav",
...     "../song.flac",
...     "sound.ogg",
...     "../files"
... ]
... 
>>> axiom_obj = Axiom(to_process, recursive = True)
>>> 
>>> results = axiom_obj.estimate(
...     start = 10,       # Begin from frame 10
...     duration = "1s",  # Analyze (max) 1 second of file
...     freq_step = 200   # Frequency drop in sample estimation
... )

Command-line console script

Simple display 🧐

user$os:~ $ # All in one, separated
user$os:~ $ axiom -i "file.mp3"
14384 7192.08984375 14 16 1 -0.001

Quiet convert 🔕

user$os:~ $ axiom -i "file.mp3" -o "file_resampled.wav" -v -1
user$os:~

Advanced ⚙️

user$os:~ $ # Start analysis from 2 seconds (-ss 2.5s) for 25000 samples (-to 25000)
user$os:~ $ # with recursive processing (-r) and verbose output (-v 2)
user$os:~ $ axiom -i "../dir" -ss 2s -to 25000 -f -r -v 1
[INFO] Progress:                        1/1 [100.00%]
[INFO] Loaded file:                     file_18400.wav
[INFO] Information:                     44100 Hz, 28680 samples (00:00.650), Mono, 56.09 KB
[INFO] ----------
[INFO] Processing samples:              0 -> 25000 (each 1) (00:00.566893) (87.17%)
[INFO] Estimated sample rate:           18841 Hz (-25259 Hz) (Linear -57.277%) (Perceptual -7.602%)
[INFO] Estimated bit depth:             10 bits (Normalized to 16 bits for safety)
[INFO] Estimated channels amount:       1 (Original: 1)
[INFO] Calculated bitrate:              301.456 kb/s
[INFO] Estimated peak:                  -0.091 dBFS
[INFO] ----------
[INFO] Loaded file:                     high_quality.flac
[INFO] Information:                     96000 Hz, 47217826 samples (08:11.852), Stereo, 88.74 MB
[INFO] ----------
[INFO] Processing samples:              192000 -> 217000 (each 1) (00:00.260417) (0.46%)
[INFO] Estimated sample rate:           48187 Hz (-47813 Hz) (Linear -49.805%) (Perceptual -0.0%)
[INFO] Estimated bit depth:             17 bits (Normalized to 32 bits for safety)
[INFO] Estimated channels amount:       2 (Original: 2)
[INFO] Calculated bitrate:              1541.984 kb/s
[INFO] Estimated peak:                  -11.207 dBFS
[INFO] ----------
[INFO] Processed 2 files in at most 00:02.772.

Disclaimer ⚠️

This software:

• Was not designed to estimate parameters of short signals.
  • Subjective measurements of FFT values related to sample rate estimation were not performed.
• Uses only vertical scanning in sample rate estimation.
• May return invalid sample rates estimations for signals with very low sample rate.
• Will return invalid bit depth estimations.