āĻ…āĻĄāĻŋāĻ“ āĻāφāχ āĻ—āĻžāχāĻĄ

RNNoise āϏāĻš āĻ¸ā§āĻĒāĻŋāϚ āĻĄāĻŋāύ⧋āχāϏāĻŋāĻ‚

RNNoise āĻšāϞ āĻāĻ•āϟāĻŋ āĻ“āĻĒ⧇āύ-āϏ⧋āĻ°ā§āϏ āύāϝāĻŧ⧇āϜ āϏāĻžāĻĒā§āϰ⧇āĻļāύ āϞāĻžāχāĻŦā§āϰ⧇āϰāĻŋ, āϝāĻž Xiph.Org āĻāĻŦāĻ‚ Mozilla-āĻāϰ āϜāĻŋāύ-āĻŽāĻžāĻ°ā§āĻ• āĻ­ā§āϝāĻžāϞāĻŋāύ āĻĻā§āĻŦāĻžāϰāĻž 2017 āϏāĻžāϞ⧇ āĻĒā§āϰāĻ•āĻžāĻļāĻŋāϤ āĻšāϝāĻŧ⧇āϛ⧇, āϝāĻž āϰāĻŋāϝāĻŧ⧇āϞ āϟāĻžāχāĻŽā§‡ āĻŦāĻ•ā§āϤ⧃āϤāĻž āĻĨ⧇āϕ⧇ āĻĒāϟāĻ­ā§‚āĻŽāĻŋāϰ āĻļāĻŦā§āĻĻ āϏāϰāĻŋāϝāĻŧ⧇ āĻĻ⧇āϝāĻŧāĨ¤

āĻāχ āύāĻŋāĻ°ā§āĻĻ⧇āĻļāĻŋāĻ•āĻž āĻļ⧁āύ⧁āύāĻāφāχ-āĻŦāĻ°ā§āĻŖāĻŋāϤ ¡ ~8 min

āĻāχ āĻĒ⧃āĻˇā§āĻ āĻžāϝāĻŧ9 āĻŽāĻŋāύāĻŋāϟ āĻĒāĻĄāĻŧāĻž
  1. āĻ“āĻ­āĻžāϰāĻ­āĻŋāω
  2. āĻŽā§‚āϞ āĻŸā§‡āĻ•āĻ“āϝāĻŧ⧇
  3. āĻ—āĻ­ā§€āϰ āĻĄā§āĻŦ
  4. RNNoise āĻĄā§‡āĻŽā§‹ āĻĻāĻŋāϝāĻŧ⧇ āĻāĻ•āϟāĻŋ āϰ⧇āĻ•āĻ°ā§āĻĄāĻŋāĻ‚ āĻĄāĻŋāύ⧋āχāϏ āĻ•āϰ⧁āύ
  5. āĻŦāĻžāĻ¸ā§āϤāĻŦ-āĻŦāĻŋāĻļā§āĻŦ āĻŦāĻžāĻ¸ā§āϤāĻŦāĻžāϝāĻŧāύ
  6. RNNoise āϏāĻš āĻŦāĻ•ā§āϤ⧃āϤāĻž āĻĄāĻŋāύ⧋āχāϏāĻŋāĻ‚āϝāĻŧ⧇āϰ āĻ­āĻŦāĻŋāĻˇā§āϝāϤ
  7. āĻāϟāĻžāϕ⧇ āĻŦāĻžāĻ¸ā§āϤāĻŦ⧇ āĻĒā§āϰāϝāĻŧā§‹āĻ— āĻ•āϰāĻž āĻšāĻšā§āϛ⧇
  8. āϏ⧂āĻ¤ā§āϰ āĻāĻŦāĻ‚ āφāϰāĻ“ āĻĒāĻĄāĻŧ⧁āύ
  9. āĻ…āĻ¨ā§āĻŦ⧇āώāĻŖ āϚāĻžāϞāĻŋāϝāĻŧ⧇ āϝāĻžāύ
  10. āĻĒā§āϰāĻžāϝāĻŧāĻļāχ āϜāĻŋāĻœā§āĻžāĻžāϏāĻŋāϤ āĻĒā§āϰāĻļā§āύāĻžāĻŦāϞ⧀

āĻ“āĻ­āĻžāϰāĻ­āĻŋāω

āĻāϟāĻŋ āĻĒā§āϰāĻžāϝāĻŧ 88,000 āĻ“āϜāύ⧇āϰ āĻāĻ•āϟāĻŋ āϛ⧋āϟ āĻĒ⧁āύāϰāĻžāĻŦ⧃āĻ¤ā§āϤ āύāĻŋāωāϰāĻžāϞ āύ⧇āϟāĻ“āϝāĻŧāĻžāĻ°ā§āϕ⧇āϰ āϏāĻžāĻĨ⧇ āĻ•ā§āϞāĻžāϏāĻŋāĻ• āϏāĻŋāĻ—āĻ¨ā§āϝāĻžāϞ āĻĒā§āϰāϏ⧇āϏāĻŋāĻ‚ āϝ⧁āĻ•ā§āϤ āĻ•āϰ⧇, āϤāĻžāχ āĻāϟāĻŋ āĻāĻ•āϟāĻŋ āϏāĻžāϧāĻžāϰāĻŖ CPU āϕ⧋āϰ⧇ āϚāϞ⧇ āĻāĻŦāĻ‚ āĻ¸ā§āĻŸā§āϰāĻŋāĻŽāĻŋāĻ‚ āĻāĻŦāĻ‚ āĻ­āϝāĻŧ⧇āϏ-āĻ•āϞ āϏāĻĢā§āϟāĻ“āϝāĻŧā§āϝāĻžāϰ⧇āϰ āĻāĻ•āϟāĻŋ āϏāĻžāϧāĻžāϰāĻŖ āĻŦāĻŋāĻ˛ā§āĻĄāĻŋāĻ‚ āĻŦā§āϞāĻ• āĻšāϝāĻŧ⧇ āĻ“āϠ⧇āĨ¤

āĻŽā§‚āϞ āĻŸā§‡āĻ•āĻ“āϝāĻŧ⧇

  1. RNNoise āĻšāϞ Xiph.Org āĻāĻŦāĻ‚ Mozilla āĻĨ⧇āϕ⧇ āĻāĻ•āϟāĻŋ BSD-āϞāĻžāχāϏ⧇āĻ¨ā§āϏāĻĒā§āϰāĻžāĻĒā§āϤ C āϞāĻžāχāĻŦā§āϰ⧇āϰāĻŋ, āϝāĻž 2017 āϏāĻžāϞ⧇ āĻĒā§āϰāĻ•āĻžāĻļāĻŋāϤ āĻšāϝāĻŧ⧇āϛ⧇, āϝāĻž āϰāĻŋāϝāĻŧ⧇āϞ āϟāĻžāχāĻŽā§‡ 48 kHz āĻ¸ā§āĻĒāĻŋāĻšā§‡āϰ āĻĒāϟāĻ­ā§‚āĻŽāĻŋāϰ āĻļāĻŦā§āĻĻāϕ⧇ āĻĻāĻŽāύ āĻ•āϰ⧇āĨ¤
  2. āĻāϟāĻŋ āĻĒā§āϰāϤāĻŋāϟāĻŋ āĻĢā§āϰāĻŋāϕ⧋āϝāĻŧ⧇āĻ¨ā§āϏāĻŋ āĻŦāĻŋāύ āĻ…āύ⧁āĻŽāĻžāύ āĻ•āϰāĻžāϰ āĻĒāϰāĻŋāĻŦāĻ°ā§āϤ⧇ āĻāĻ•āϟāĻŋ āϛ⧋āϟ āϗ⧇āĻŸā§‡āĻĄ āĻĒ⧁āύāϰāĻžāĻŦ⧃āĻ¤ā§āϤ āύ⧇āϟāĻ“āϝāĻŧāĻžāĻ°ā§āϕ⧇āϰ āϏāĻžāĻĨ⧇ 22 āĻŦā§āϝāĻžāĻ¨ā§āĻĄ āϞāĻžāϭ⧇āϰ āĻĒā§‚āĻ°ā§āĻŦāĻžāĻ­āĻžāϏ āĻĻ⧇āϝāĻŧ, āϝāĻž āĻ—āĻŖāύāĻž āĻ•āĻŽ āϰāĻžāϖ⧇āĨ¤
  3. 2018 āĻĒ⧇āĻĒāĻžāϰ⧇ 87,503 āĻ“āϜāύ āĻāĻŦāĻ‚ āĻāĻ•āϟāĻŋ x86 āϕ⧋āϰ⧇āϰ āĻĒā§āϰāĻžāϝāĻŧ 1.3 āĻļāϤāĻžāĻ‚āĻļ, āϕ⧋āύ GPU āĻĒā§āϰāϝāĻŧā§‹āϜāύ āĻ›āĻžāĻĄāĻŧāĻžāχ āϰāĻŋāĻĒā§‹āĻ°ā§āϟ āĻ•āϰāĻž āĻšāϝāĻŧ⧇āϛ⧇āĨ¤
  4. āĻāĻ•āϟāĻŋ āĻĒāĻŋāϚ-āϟāĻŋāωāύāϝ⧁āĻ•ā§āϤ āϚāĻŋāϰ⧁āύāĻŋ āĻĢāĻŋāĻ˛ā§āϟāĻžāϰ āĻ­āϝāĻŧ⧇āϏ āĻšāĻžāϰāĻŽā§‹āύāĻŋāĻ•ā§āϏ⧇āϰ āĻŽāĻ§ā§āϝ⧇ āĻ—ā§‹āϞāĻŽāĻžāϞ āϏāϰāĻŋāϝāĻŧ⧇ āĻĻ⧇āϝāĻŧ āϝāĻž āĻŽā§‹āϟāĻž āĻŦā§āϝāĻžāĻ¨ā§āĻĄ āϞāĻžāĻ­ āĻ•āϰāϤ⧇ āĻĒāĻžāϰ⧇ āύāĻžāĨ¤
  5. OBS āĻ¸ā§āϟ⧁āĻĄāĻŋāĻ“ 2020 āϏāĻžāϞ⧇āϰ āϏ⧇āĻĒā§āĻŸā§‡āĻŽā§āĻŦāϰ⧇ 26.0 āϏāĻ‚āĻ¸ā§āĻ•āϰāϪ⧇ āĻāĻ•āϟāĻŋ āĻļāĻŦā§āĻĻ āĻĻāĻŽāύ āĻĒāĻĻā§āϧāϤāĻŋ āĻšāĻŋāϏāĻžāĻŦ⧇ RNNoise āϝ⧁āĻ•ā§āϤ āĻ•āϰ⧇āϛ⧇āĨ¤

āĻ—āĻ­ā§€āϰ āĻĄā§āĻŦ

RNNoise is a speech enhancement library: software that takes audio containing someone talking over background noise, such as a fan, traffic or keyboard clicks, and returns a cleaner version in which the voice is kept and the noise is reduced. It was written by Jean-Marc Valin and published in September 2017 as a Xiph.Org and Mozilla project with a browser demo and a technical paper. The paper, A Hybrid DSP/Deep Learning Approach to Real-Time Full-Band Speech Enhancement, appeared at the 2018 IEEE Workshop on Multimedia Signal Processing. The code is written in C and distributed under a BSD license.

The design goal was practical: run in real time on modest hardware at full-band 48 kHz audio, the sample rate used in high-quality voice calls, without a GPU. Many deep learning denoisers of the time estimated every frequency bin of a spectrogram, or even every audio sample, which takes large networks. RNNoise instead keeps the parts of a traditional noise suppressor that signal processing already does well, such as framing, spectral analysis and pitch tracking, and asks a neural network to do only the hardest part, which is estimating how much each part of the spectrum should be turned down.

The processing happens in short, overlapping frames. Audio is cut into 20 millisecond windows that advance every 10 milliseconds, and each window is converted to the frequency domain with a Fourier transform. Rather than handle hundreds of individual frequency bins, RNNoise groups them into 22 bands spaced on the Bark scale, a scale that roughly matches how human hearing resolves frequency, using the band layout of the Opus codec. For each frame it computes 42 input features: 22 Bark-frequency cepstral coefficients that summarize the spectral shape, the first and second time derivatives of the first six, six coefficients describing how strongly each band follows the voice's pitch, the pitch period itself and a measure of how stationary the spectrum is.

Those features go into a recurrent neural network, a network that keeps an internal state from frame to frame so it can track how sounds evolve over time. The network uses gated recurrent units (GRUs), a type of recurrent layer that learns what to remember and what to forget, combined with fully connected layers. The paper reports four hidden layers with 215 units in total, the largest having 96, and 87,503 weights. The network produces 22 gains between 0 and 1, one per band, plus a voice activity probability. Each gain multiplies the energy in its band: a value near 1 lets the band through, near 0 suppresses it. Because 22 band gains are too coarse to remove noise sitting between the harmonics of a voice, a final pitch filter, a comb filter tuned to the detected pitch, attenuates energy between harmonics. The result is converted back to a waveform by overlap-add.

Training uses synthetic mixtures. Clean speech and noise are rarely recorded together with a clean reference, so they are recorded separately and mixed at random levels with random filtering. The paper reports about 6 hours of speech and 4 hours of noise expanded this way into about 140 hours of noisy training audio. The target for each band is the square root of the ratio between clean speech energy and noisy energy, and the loss compares gains after a further square root, which penalizes mistakes on quiet bands more. Because the system does not normalize levels, the training data must include speech at every realistic volume.

The efficiency numbers explain its adoption. The paper reports that the network needs about 175,000 floating point operations per frame, or 17.5 million per second at 100 frames per second, and about 40 million per second for the whole algorithm including transforms and pitch search. That used about 1.3 percent of one core of an Intel Haswell laptop processor and about 14 percent of a Raspberry Pi 3 core. Weights constrained to plus or minus 0.5 during training can be stored as 8-bit values, giving a model of about 85 kB according to the demo page. In a PESQ evaluation, an objective speech quality score, the paper reports clearly better quality than the conventional suppressor in SpeexDSP, particularly for non-stationary noise such as babble, car and street noise.

Compared with the alternatives, RNNoise sits between old and new. Classical suppressors such as the Speex preprocessor estimate the noise spectrum statistically and subtract it; they are cheap but struggle with noise that changes quickly. Large neural denoisers trained for the Microsoft-run Deep Noise Suppression Challenge at INTERSPEECH 2020 can remove more noise but cost more compute; the challenge organizers ranked entries with crowdsourced listening tests under the ITU-T P.808 standard because objective metrics often disagreed with human judgments. Valin's own follow-up with co-authors, PercepNet, presented at INTERSPEECH 2020, kept the band-and-pitch design but reported higher quality while still using less than 5 percent of a CPU core.

The limitations follow from the design. A band gain cannot separate a voice from noise that occupies the same band at the same moment, so strong noise leaves a muffled or watery voice, a distortion often called musical noise. RNNoise is trained to separate speech from background sounds, not one talker from another, so a competing voice in the background is treated as speech. Its quality is bounded by the training mixtures, and a 2017-era model can misjudge noise types it never heard. The 10 millisecond look-ahead keeps latency low for calls but limits how much future context the network can use.

Using RNNoise in practice requires little. The repository builds with a standard configure and make process, downloads the model files during setup, and includes a demo that reads headerless 16-bit mono audio at 48 kHz. Developers who want a custom model can generate features from their own speech and noise files with the provided tools, train with the included Python script and compile the quantized weights back into the library. The README recommends at least 10,000 training sequences and preferably 200,000 or more.

04āĻ•āĻžāĻœā§‡āϰ āωāĻĻāĻžāĻšāϰāĻŖ

RNNoise āĻĄā§‡āĻŽā§‹ āĻĻāĻŋāϝāĻŧ⧇ āĻāĻ•āϟāĻŋ āϰ⧇āĻ•āĻ°ā§āĻĄāĻŋāĻ‚ āĻĄāĻŋāύ⧋āχāϏ āĻ•āϰ⧁āύ

  1. āĻāĻ•āϟāĻŋ āϚāϞāĻŽāĻžāύ āĻĢā§āϝāĻžāύ⧇āϰ āĻ•āĻžāϛ⧇ āĻĒā§āϰāĻžāϝāĻŧ 30 āϏ⧇āϕ⧇āĻ¨ā§āĻĄā§‡āϰ āĻŦāĻ•ā§āϤ⧃āϤāĻž āϰ⧇āĻ•āĻ°ā§āĻĄ āĻ•āϰ⧁āύ, āĻŦāĻž āĻāĻ•āϟāĻŋ āĻļāĻŦā§āĻĻ āĻ•ā§āϞāĻŋāĻĒ⧇āϰ āϏāĻžāĻĨ⧇ āĻĒāϰāĻŋāĻˇā§āĻ•āĻžāϰ āĻŦāĻ•ā§āϤ⧃āϤāĻž āĻŽāĻŋāĻļā§āϰāĻŋāϤ āĻ•āϰ⧁āύāĨ¤

  2. āĻĢāĻžāχāϞāϟāĻŋāϕ⧇ 48 kHz-āĻ āĻ•āĻžāρāϚāĻž āĻšā§‡āĻĄāĻžāϰāϞ⧇āϏ 16-āĻŦāĻŋāϟ āĻŽāύ⧋ PCM-āĻ āϰ⧂āĻĒāĻžāĻ¨ā§āϤāϰ āĻ•āϰ⧁āύ, āωāĻĻāĻžāĻšāϰāĻŖāĻ¸ā§āĻŦāϰ⧂āĻĒ āĻāĻ•āϟāĻŋ āĻ•āĻŽāĻžāĻ¨ā§āĻĄ-āϞāĻžāχāύ āĻ…āĻĄāĻŋāĻ“ āĻ•āύāĻ­āĻžāĻ°ā§āϟāĻžāϰ āϏāĻšāĨ¤

  3. Autogen.sh āĻāϰ āϏāĻžāĻĨ⧇ āĻ…āĻĢāĻŋāϏāĻŋāϝāĻŧāĻžāϞ āϰāĻŋāĻĒā§‹āϜāĻŋāϟāϰāĻŋ āĻĨ⧇āϕ⧇ RNNoise āϤ⧈āϰāĻŋ āĻ•āϰ⧁āύ, āĻ•āύāĻĢāĻŋāĻ—āĻžāϰ āĻ•āϰ⧁āύ āĻāĻŦāĻ‚ āϤ⧈āϰāĻŋ āĻ•āϰ⧁āύ, āϝāĻž āĻŽāĻĄā§‡āϞ āĻĢāĻžāχāϞāϗ⧁āϞāĻŋāĻ“ āĻĄāĻžāωāύāϞ⧋āĻĄ āĻ•āϰ⧇āĨ¤

  4. āχāύāĻĒ⧁āϟ āĻšāĻŋāϏāĻžāĻŦ⧇ āĻļā§‹āϰāĻ—ā§‹āϞ āĻĢāĻžāχāϞ āĻāĻŦāĻ‚ āφāωāϟāĻĒ⧁āϟ āĻšāĻŋāϏāĻžāĻŦ⧇ āĻāĻ•āϟāĻŋ āύāϤ⧁āύ āĻĢāĻžāχāϞ⧇āϰ āύāĻžāĻŽ āϏāĻš āωāĻĻāĻžāĻšāϰāĻŖ/rnnoise_demo āϚāĻžāϞāĻžāύāĨ¤

  5. āφāωāϟāĻĒ⧁āϟāϟāĻŋāϕ⧇ āφāĻŦāĻžāϰ WAV-āϤ⧇ āϰ⧂āĻĒāĻžāĻ¨ā§āϤāϰ āĻ•āϰ⧁āύ āĻāĻŦāĻ‚ āφāϏāϞ āĻāĻŦāĻ‚ āĻ…āĻ¸ā§āĻŦā§€āĻ•ā§ƒāϤ āϏāĻ‚āĻ¸ā§āĻ•āϰāĻŖāϗ⧁āϞāĻŋ āĻĒāĻŋāĻ›āύ⧇āϰ āĻĻāĻŋāϕ⧇ āĻļ⧁āύ⧁āύāĨ¤

  6. āĻāĻ•āϟāĻŋ āĻ¸ā§āĻĒ⧇āĻ•āĻŸā§āϰ⧋āĻ—ā§āϰāĻžāĻŽ āĻ­āĻŋāωāϝāĻŧāĻžāϰ⧇ āωāĻ­āϝāĻŧ āĻĢāĻžāχāϞ āϖ⧁āϞ⧁āύ āĻāĻŦāĻ‚ āĻ¸ā§āĻĒāĻŋāϚ āĻšāĻžāϰāĻŽā§‹āύāĻŋāĻ•ā§āϏ āĻāĻŦāĻ‚ āĻŦāĻŋāϰāϤāĻŋāϰ āϏāĻŽāϝāĻŧ āĻļāĻ•ā§āϤāĻŋāϰ āϤ⧁āϞāύāĻž āĻ•āϰ⧁āύāĨ¤

āĻāϟāĻž āĻ•āĻŋ āĻĻ⧇āĻ–āĻžāϝāĻŧ

āϞāĻžāĻ°ā§āύāĻžāϰ āĻļ⧇āώ āĻšāϝāĻŧ āĻāĻ•āϟāĻŋ āĻĄāĻŋāύ⧋āχāϜāĻĄ āĻĢāĻžāχāϞ āĻĻāĻŋāϝāĻŧ⧇ āϝ⧇āĻ–āĻžāύ⧇ āĻ¸ā§āĻĨāĻŋāϰ āĻĢā§āϝāĻžāύ⧇āϰ āφāĻ“āϝāĻŧāĻžāϜ āĻĻ⧃āĻĸāĻŧāĻ­āĻžāĻŦ⧇ āĻ•āĻŽā§‡ āϝāĻžāϝāĻŧ, āĻŦāĻŋāĻļ⧇āώ āĻ•āϰ⧇ āĻŦāĻŋāϰāϤāĻŋāϤ⧇āĨ¤ āĻ¸ā§āĻĒ⧇āĻ•āĻŸā§āϰ⧋āĻ—ā§āϰāĻžāĻŽāϟāĻŋ āĻ­āϝāĻŧ⧇āϏ āĻšāĻžāϰāĻŽā§‹āύāĻŋāĻ•ā§āϏ⧇āϰ āĻŽāĻ§ā§āϝ⧇ āĻ—āĻžāĻĸāĻŧ āĻŦā§āϝāĻŦāϧāĻžāύ āĻĻ⧇āĻ–āĻžāϤ⧇ āĻšāĻŦ⧇, āϝāĻ–āύ āϖ⧁āĻŦ āĻœā§‹āϰ⧇ āφāĻ“āϝāĻŧāĻžāϜ āĻ­āϝāĻŧ⧇āϏāϕ⧇ āĻ“āĻ­āĻžāϰāĻ˛ā§āϝāĻžāĻĒ āĻ•āϰ⧇ āĻ•āĻŋāϛ⧁ āϧāĻžāĻ•ā§āĻ•āĻž āĻĻāĻŋāϤ⧇ āĻĒāĻžāϰ⧇, āĻŦā§āϝāĻžāĻ¨ā§āĻĄ-āϞāĻžāϭ⧇āϰ āϏ⧀āĻŽāĻžāĻŦāĻĻā§āϧāϤāĻžāϕ⧇ āϚāĻŋāĻ¤ā§āϰāĻŋāϤ āĻ•āϰ⧇āĨ¤

āĻŦāĻžāĻ¸ā§āϤāĻŦ-āĻŦāĻŋāĻļā§āĻŦ āĻŦāĻžāĻ¸ā§āϤāĻŦāĻžāϝāĻŧāύ

OBS āĻ¸ā§āϟ⧁āĻĄāĻŋāĻ“ 26.0, āϏ⧇āĻĒā§āĻŸā§‡āĻŽā§āĻŦāϰ 2020 āĻ āĻĒā§āϰāĻ•āĻžāĻļāĻŋāϤ āĻšāϝāĻŧ⧇āϛ⧇, RNNoise āĻāϰ āύāϝāĻŧ⧇āϜ āϏāĻžāĻĒā§āϰ⧇āĻļāύ āĻĢāĻŋāĻ˛ā§āϟāĻžāϰ⧇ āĻāĻ•āϟāĻŋ āĻĒāĻĻā§āϧāϤāĻŋ āĻšāĻŋāϏāĻžāĻŦ⧇ āϝ⧁āĻ•ā§āϤ āĻ•āϰ⧇āϛ⧇, āϝāĻž āĻ¸ā§āĻŸā§āϰāĻŋāĻŽāĻžāϰāĻĻ⧇āϰ āĻĒ⧁āϰāĻžāύ⧋ āĻ¸ā§āĻĒā§€āĻ•ā§āϏ-āĻ­āĻŋāĻ¤ā§āϤāĻŋāĻ• āĻĻāĻŽāύāĻ•āĻžāϰ⧀āϰ āĻĒāĻžāĻļāĻžāĻĒāĻžāĻļāĻŋ āĻāĻ•āϟāĻŋ āύāĻŋāωāϰāĻžāϞ āĻŦāĻŋāĻ•āĻ˛ā§āĻĒ āĻĻāĻŋāϝāĻŧ⧇āϛ⧇āĨ¤

GamingOnLinux āϰāĻŋāĻĒā§‹āĻ°ā§āϟ āĻ•āϰ⧇āϛ⧇ āϝ⧇ āϞāĻŋāύāĻžāĻ•ā§āϏ āϟ⧁āϞāϏ NoiseTorch āĻāĻŦāĻ‚ Cadmus āĻāĻ•āχ RNNoise āĻĒā§āϰāϝ⧁āĻ•ā§āϤāĻŋ āĻŦā§āϝāĻŦāĻšāĻžāϰ āĻ•āϰ⧇ āĻ­āϝāĻŧ⧇āϏ āĻšā§āϝāĻžāϟ āĻāĻŦāĻ‚ āĻ•āϞ⧇āϰ āϜāĻ¨ā§āϝ āĻŽāĻžāχāĻ•ā§āϰ⧋āĻĢā§‹āύ āχāύāĻĒ⧁āϟ āϏāĻŋāĻ¸ā§āĻŸā§‡āĻŽ-āĻŦā§āϝāĻžāĻĒā§€ āĻĒāϰāĻŋāĻˇā§āĻ•āĻžāϰ āĻ•āϰāϤ⧇āĨ¤

āĻ…āĻĢāĻŋāϏāĻŋāϝāĻŧāĻžāϞ āϰāĻŋāĻĒā§‹āϜāĻŋāϟāϰāĻŋāϤ⧇ rnnoise_demo āĻĒā§āϰ⧋āĻ—ā§āϰāĻžāĻŽāϟāĻŋ āĻ•āĻŽāĻžāĻ¨ā§āĻĄ āϞāĻžāχāύ āĻĨ⧇āϕ⧇ āĻāĻ•āϟāĻŋ āĻ•āĻžāρāϚāĻž 16-āĻŦāĻŋāϟ, 48 kHz āĻŽāύ⧋ āϰ⧇āĻ•āĻ°ā§āĻĄāĻŋāĻ‚āϕ⧇ āĻ…āĻ¸ā§āĻŦā§€āĻ•āĻžāϰ āĻ•āϰ⧇, āϝāĻž āĻĒāĻĄāĻ•āĻžāĻ¸ā§āϟ āĻŦāĻž āĻŽāĻŋāϟāĻŋāĻ‚ āĻ…āĻĄāĻŋāĻ“āϤ⧇ āĻĒāϰ⧀āĻ•ā§āώāĻž āĻ•āϰāĻž āϏāĻšāϜ āĻ•āϰ⧇ āϤ⧋āϞ⧇āĨ¤

āĻ­ā§āϝāĻžāϞāĻŋāύ⧇āϰ āύāϝāĻŧ⧇āϜ āĻĄā§‹āύ⧇āĻļāύ āĻĒā§āϰāĻ•āĻ˛ā§āĻĒ āϜāύāϏāĻžāϧāĻžāϰāĻŖāϕ⧇ āϤāĻžāĻĻ⧇āϰ āύāĻŋāϜāĻ¸ā§āĻŦ āĻĒāϰāĻŋāĻŦ⧇āĻļ⧇āϰ āĻāĻ• āĻŽāĻŋāύāĻŋāĻŸā§‡āϰ āĻļāĻŦā§āĻĻ āϰ⧇āĻ•āĻ°ā§āĻĄ āĻ•āϰāĻžāϰ āϜāĻ¨ā§āϝ āφāĻŽāĻ¨ā§āĻ¤ā§āϰāĻŖ āϜāĻžāύāĻŋāϝāĻŧ⧇āϛ⧇ āĻāĻŦāĻ‚ āϏāĻ‚āĻ—ā§ƒāĻšā§€āϤ āϰ⧇āĻ•āĻ°ā§āĻĄāĻŋāĻ‚āϗ⧁āϞāĻŋ āĻĒā§āϰāĻļāĻŋāĻ•ā§āώāϪ⧇āϰ āϜāĻ¨ā§āϝ āĻāĻ•āϟāĻŋ 6.4 GB āĻĄāĻžāωāύāϞ⧋āĻĄāϝ⧋āĻ—ā§āϝ āϏāĻ‚āϰāĻ•ā§āώāĻŖāĻžāĻ—āĻžāϰ āĻšāĻŋāϏāĻžāĻŦ⧇ āĻĒā§āϰāĻ•āĻžāĻļ āĻ•āϰāĻž āĻšāϝāĻŧ⧇āĻ›āĻŋāϞāĨ¤

RNNoise 0.2, āĻāĻĒā§āϰāĻŋāϞ 2024-āĻ āĻĒā§āϰāĻ•āĻžāĻļāĻŋāϤ, āĻļāĻŋāĻĒāĻĄ āĻŽāĻĄā§‡āϞāϗ⧁āϞāĻŋāϕ⧇ āĻļ⧁āϧ⧁āĻŽāĻžāĻ¤ā§āϰ āϏāĻ°ā§āĻŦāϜāύ⧀āύāĻ­āĻžāĻŦ⧇ āωāĻĒāϞāĻŦā§āϧ āĻĄā§‡āϟāĻžāϏ⧇āĻŸā§‡ āĻĒ⧁āύāϰāĻžāϝāĻŧ āĻĒā§āϰāĻļāĻŋāĻ•ā§āώāĻŖ āĻĻ⧇āĻ“āϝāĻŧāĻž āĻšāϝāĻŧ⧇āϛ⧇ āĻāĻŦāĻ‚ āϰāĻžāύ-āϟāĻžāχāĻŽ CPU āϏāύāĻžāĻ•ā§āϤāĻ•āϰāϪ⧇āϰ āϏāĻžāĻĨ⧇ AVX2 āĻāĻŦāĻ‚ SSE4.1 āĻ…āĻĒā§āϟāĻŋāĻŽāĻžāχāĻœā§‡āĻļāĻžāύ āϝ⧋āĻ— āĻ•āϰāĻž āĻšāϝāĻŧ⧇āϛ⧇āĨ¤

RNNoise āϏāĻš āĻŦāĻ•ā§āϤ⧃āϤāĻž āĻĄāĻŋāύ⧋āχāϏāĻŋāĻ‚āϝāĻŧ⧇āϰ āĻ­āĻŦāĻŋāĻˇā§āϝāϤ

RNNoise-āĻāϰ 2024 āϰāĻŋāϞāĻŋāϜ, āϏāĻ‚āĻ¸ā§āĻ•āϰāĻŖ 0.2, āĻŽā§‚āϞ āϧāĻžāϰāĻŖāĻžāϰ āĻĒāϰāĻŋāĻŦāĻ°ā§āϤ⧇ āĻĒā§āϰāĻļāĻŋāĻ•ā§āώāĻŖ āĻāĻŦāĻ‚ āĻĒā§āϰāĻ•ā§ŒāĻļāϞ āĻĒāϰāĻŋāĻŦāĻ°ā§āϤāύ āĻ•āϰ⧇āϛ⧇: Phoronix āĻāĻĒā§āϰāĻŋāϞ 2024 āĻ āϰāĻŋāĻĒā§‹āĻ°ā§āϟ āĻ•āϰ⧇āϛ⧇ āϝ⧇ āĻŦāĻŋāϤāϰāĻŖ āĻ•āϰāĻž āĻŽāĻĄā§‡āϞāϗ⧁āϞāĻŋ āĻļ⧁āϧ⧁āĻŽāĻžāĻ¤ā§āϰ āϏāĻ°ā§āĻŦāϜāύ⧀āύāĻ­āĻžāĻŦ⧇ āωāĻĒāϞāĻŦā§āϧ āĻĄā§‡āϟāĻžāϏ⧇āϟ āĻŦā§āϝāĻŦāĻšāĻžāϰ āĻ•āϰ⧇ āĻĒ⧁āύāϰāĻžāϝāĻŧ āĻĒā§āϰāĻļāĻŋāĻ•ā§āώāĻŖ āĻĻ⧇āĻ“āϝāĻŧāĻž āĻšāϝāĻŧ⧇āϛ⧇ āĻāĻŦāĻ‚ āϕ⧋āĻĄāϟāĻŋ āϰāĻžāύ-āϟāĻžāχāĻŽ CPU āϏāύāĻžāĻ•ā§āϤāĻ•āϰāϪ⧇āϰ āϏāĻžāĻĨ⧇ SSE4.1 āĻāĻŦāĻ‚ AVX2 āĻ…āĻĒā§āϟāĻŋāĻŽāĻžāχāĻœā§‡āĻļāĻžāύ āĻ…āĻ°ā§āϜāύ āĻ•āϰ⧇āϛ⧇āĨ¤ āϏāĻ‚āĻ—ā§āϰāĻšāĻ¸ā§āĻĨāϞāϟāĻŋ āĻāĻ–āύ āĻāĻ•āϟāĻŋ āĻ¸ā§āĻĒāĻžāĻ°ā§āϏāĻžāϰ āϛ⧋āϟ āĻŽāĻĄā§‡āϞ āĻāĻŦāĻ‚ āϰāĻžāύ āϟāĻžāχāĻŽā§‡ āĻāĻ•āϟāĻŋ āĻĢāĻžāχāϞ āĻĨ⧇āϕ⧇ āĻ“āϜāύ āϞ⧋āĻĄ āĻ•āϰāĻžāϰ āĻāĻ•āϟāĻŋ āωāĻĒāĻžāϝāĻŧ āĻ…āĻĢāĻžāϰ āĻ•āϰ⧇, āϝāĻž āĻ•āĻžāĻ¸ā§āϟāĻŽ āĻŽāĻĄā§‡āϞāϗ⧁āϞāĻŋāϤ⧇ āĻ…āĻĻāϞāĻŦāĻĻāϞ āĻ•āϰāĻž āϏāĻšāϜ āĻ•āϰ⧇ āϤ⧋āϞ⧇āĨ¤ āĻŦāĻŋāĻ¸ā§āϤ⧃āϤ āĻ—āĻŦ⧇āώāĻŖāĻž āĻĻāĻŋāĻ•, PercepNet (INTERSPEECH 2020) āĻāĻŦāĻ‚ 2020 āϏāĻžāϞ⧇ āĻļ⧁āϰ⧁ āĻšāĻ“āϝāĻŧāĻž āĻĄāĻŋāĻĒ āύāϝāĻŧ⧇āϜ āϏāĻžāĻĒā§āϰ⧇āĻļāύ āĻšā§āϝāĻžāϞ⧇āĻžā§āϜ āϏāĻŋāϰāĻŋāĻœā§‡ āĻĻ⧃āĻļā§āϝāĻŽāĻžāύ, āĻ…āύ⧁āϰ⧂āĻĒ āĻ•āĻŽ āĻ•āĻŽā§āĻĒāĻŋāωāĻŸā§‡ āωāĻšā§āϚāϤāϰ āĻ…āύ⧁āĻ­ā§‚āϤ āĻŽāĻžāύ⧇āϰ āĻĻāĻŋāϕ⧇, āϝāĻž āĻļ⧁āϧ⧁āĻŽāĻžāĻ¤ā§āϰ āωāĻĻā§āĻĻ⧇āĻļā§āϝāĻŽā§‚āϞāĻ• āĻ¸ā§āϕ⧋āϰ⧇āϰ āĻšā§‡āϝāĻŧ⧇ āĻŽāĻžāύ⧁āώ⧇āϰ āĻļā§āϰāĻŦāĻŖ āĻĒāϰ⧀āĻ•ā§āώāĻž āĻĻā§āĻŦāĻžāϰāĻž āĻŦāĻŋāϚāĻžāϰ āĻ•āϰāĻž āĻšāϝāĻŧ, āϝ⧇āĻšā§‡āϤ⧁ āĻšā§āϝāĻžāϞ⧇āĻžā§āĻœā§‡āϰ āφāϝāĻŧā§‹āϜāĻ•āϰāĻž āĻĒā§āϰāĻžāϝāĻŧāĻļāχ āĻāχ āĻĻ⧁āϟāĻŋāϰ āϏāĻžāĻĨ⧇ āĻāĻ•āĻŽāϤ āĻšāύ āύāĻžāĨ¤ āωāĻ¨ā§āĻŽā§āĻ•ā§āϤ āϏāĻŽāĻ¸ā§āϝāĻžāϗ⧁āϞāĻŋāϰ āĻŽāĻ§ā§āϝ⧇ āϰāϝāĻŧ⧇āϛ⧇ āĻĒā§āϰāϤāĻŋāϝ⧋āĻ—ā§€ āĻŦāĻ•ā§āϤāĻžāĻĻ⧇āϰ āĻĻāĻŽāύ āĻ•āϰāĻž āĻāĻŦāĻ‚ āϟāĻžāĻ°ā§āϗ⧇āϟ āĻ­āϝāĻŧ⧇āϏ⧇āϰ āĻ•ā§āώāϤāĻŋ āύāĻž āĻ•āϰ⧇āχ āĻ§ā§āĻŦāύāĻŋāϤ āĻ•āϰāĻž, āĻĒā§āϰāĻļāĻŋāĻ•ā§āώāϪ⧇āϰ āĻĄā§‡āϟāĻž āĻĨ⧇āϕ⧇ āĻ…āύ⧁āĻĒāĻ¸ā§āĻĨāĻŋāϤ āĻļāĻŦā§āĻĻ⧇āϰ āĻĒā§āϰāĻ•āĻžāϰ⧇āϰ āĻĻ⧃āĻĸāĻŧāϤāĻž, āĻāĻŦāĻ‚ āĻŽā§‚āĻ˛ā§āϝāĻžāϝāĻŧāύ āϝāĻž āϏāĻŋāĻ¨ā§āĻĨ⧇āϟāĻŋāĻ• āĻŽāĻŋāĻļā§āϰāϪ⧇āϰ āĻĒāϰāĻŋāĻŦāĻ°ā§āϤ⧇ āĻŦāĻžāĻ¸ā§āϤāĻŦ āĻ•āϞāϗ⧁āϞāĻŋāϕ⧇ āĻĒā§āϰāϤāĻŋāĻĢāϞāĻŋāϤ āĻ•āϰ⧇āĨ¤ RNNoise-āĻāϰ āĻŽāϤ⧋ āϛ⧋āϟ āĻšāĻžāχāĻŦā§āϰāĻŋāĻĄ āĻĄāĻŋāϜāĻžāχāύ āĻŦāĻž āĻĄā§‡āĻĄāĻŋāϕ⧇āĻŸā§‡āĻĄ āĻšāĻžāĻ°ā§āĻĄāĻ“āϝāĻŧā§āϝāĻžāϰ⧇ āϚāϞāĻŽāĻžāύ āĻŦ⧃āĻšāĻ¤ā§āϤāϰ āϏāĻŽā§āĻĒā§‚āĻ°ā§āĻŖ āύāĻŋāωāϰāĻžāϞ āĻŽāĻĄā§‡āϞāϗ⧁āϞāĻŋ āĻ­āϝāĻŧ⧇āϏ āĻ•āϞ⧇āϰ āϜāĻ¨ā§āϝ āĻĄāĻŋāĻĢāĻ˛ā§āϟ āĻšāϝāĻŧ⧇ āωāĻ āĻŦ⧇ āĻ•āĻŋāύāĻž āϤāĻž āĻāĻ•āϟāĻŋ āωāĻ¨ā§āĻŽā§āĻ•ā§āϤ āĻĒā§āϰāĻļā§āύ āĻĨ⧇āϕ⧇ āϝāĻžāϝāĻŧāĨ¤

āĻāϟāĻžāϕ⧇ āĻŦāĻžāĻ¸ā§āϤāĻŦ⧇ āĻĒā§āϰāϝāĻŧā§‹āĻ— āĻ•āϰāĻž āĻšāĻšā§āϛ⧇

āϕ⧇āύ āĻāϟāĻž āϗ⧁āϰ⧁āĻ¤ā§āĻŦāĻĒā§‚āĻ°ā§āĻŖ

  • āĻ…ā§āϝāĻžāĻ•ā§āϏ⧇āϏ āĻāĻŦāĻ‚ āĻĒ⧌āρāĻ›āĻžāύ⧋. āĻāϟāĻŋ āĻŸā§āϰāĻžāĻ¨ā§āϏāĻ•ā§āϰāĻŋāĻĒāĻļāύ, āĻŦāĻ°ā§āĻŖāύāĻž āĻāĻŦāĻ‚ āĻ­āϝāĻŧ⧇āϏ āχāĻ¨ā§āϟāĻžāϰāĻĢ⧇āϏ⧇āϰ āĻŽāĻžāĻ§ā§āϝāĻŽā§‡ āĻ…ā§āϝāĻžāĻ•ā§āϏ⧇āϏāϝ⧋āĻ—ā§āϝāϤāĻž āωāĻ¨ā§āύāϤ āĻ•āϰ⧇āĨ¤
  • āĻ–āϰāϚ āĻ“ āĻŦāĻžāĻœā§‡āϟ. āĻŽāĻŋāĻĄāĻŋāϝāĻŧāĻž āĻĻāϞāϗ⧁āϞāĻŋ āϛ⧋āϟ āĻŦāĻžāĻœā§‡āĻŸā§‡āϰ āϏāĻžāĻĨ⧇ āĻĒāĻžāϞāĻŋāĻļ āĻ•āϰāĻž āĻ…āĻĄāĻŋāĻ“ āĻĻā§āϰ⧁āϤ āĻĒāĻžāĻ āĻžāϤ⧇ āĻĒāĻžāϰ⧇āĨ¤
  • āĻ—āϤāĻŋ āĻāĻŦāĻ‚ āĻ¸ā§āϕ⧇āϞ. āĻ—ā§āϰāĻžāĻšāĻ•-āĻŽā§āĻ–ā§€ āϏāĻŋāĻ¸ā§āĻŸā§‡āĻŽāϗ⧁āϞāĻŋ āĻŦ⧃āĻšāĻ¤ā§āϤāϰ āĻ¸ā§āϕ⧇āϞ⧇ āĻ•āĻĨā§āϝ āĻŽāĻŋāĻĨāĻ¸ā§āĻ•ā§āϰāĻŋāϝāĻŧāĻž āĻĒā§āϰāĻ•ā§āϰāĻŋāϝāĻŧāĻž āĻ•āϰāϤ⧇ āĻĒāĻžāϰ⧇āĨ¤

āϜāĻ¨ā§āϝ āϏāϤāĻ°ā§āĻ•

  • āϏāĻŽā§āĻŽāϤāĻŋ āĻ…āύ⧁āĻĒāĻ¸ā§āĻĨāĻŋāϤ āĻšāϞ⧇ āĻ­āϝāĻŧ⧇āϏ āĻ…āĻĒāĻŦā§āϝāĻŦāĻšāĻžāϰ āĻāĻŦāĻ‚ āĻ›āĻĻā§āĻŽāĻŦ⧇āĻļ⧇āϰ āĻā§āρāĻ•āĻŋ āĻŦ⧇āĻĄāĻŧ⧇ āϝāĻžāϝāĻŧāĨ¤
  • āωāĻšā§āϚāĻžāϰāĻŖ, āωāĻĒāĻ­āĻžāώāĻž āĻŦāĻž āϕ⧋āϞāĻžāĻšāϞāĻĒā§‚āĻ°ā§āĻŖ āĻĒāϰāĻŋāĻŦ⧇āĻļ⧇ āϝāĻĨāĻžāĻ°ā§āĻĨāϤāĻž āĻšā§āϰāĻžāϏ āĻĒ⧇āϤ⧇ āĻĒāĻžāϰ⧇āĨ¤
  • āϏāĻŋāĻ¨ā§āĻĨ⧇āϟāĻŋāĻ• āĻ…āĻĄāĻŋāĻ“ āĻĒāϰāĻŋāĻˇā§āĻ•āĻžāϰ āϞ⧇āĻŦ⧇āϞ āĻ›āĻžāĻĄāĻŧāĻž āĻ–āĻžāρāϟāĻŋ āĻŦāĻ•ā§āϤ⧃āϤāĻž āĻšāĻŋāϏāĻžāĻŦ⧇ āϭ⧁āϞ āĻšāϤ⧇ āĻĒāĻžāϰ⧇āĨ¤

āĻĒāϰāĻŦāĻ°ā§āϤ⧀ āĻĒāĻĻāĻ•ā§āώ⧇āĻĒ

  1. āĻ­āϝāĻŧ⧇āϏ āĻ•ā§āϝāĻžāĻĒāϚāĻžāϰ, āĻ•ā§āϞ⧋āύāĻŋāĻ‚ āĻāĻŦāĻ‚ āĻĒ⧁āύāσāĻŦā§āϝāĻŦāĻšāĻžāϰ⧇āϰ āϜāĻ¨ā§āϝ āϏ⧁āĻ¸ā§āĻĒāĻˇā§āϟ āϏāĻŽā§āĻŽāϤāĻŋ āĻĒāĻžāύāĨ¤
  2. āĻŦāĻŋāĻ­āĻŋāĻ¨ā§āύ āĻ¸ā§āĻĒāĻŋāĻ•āĻžāϰ āĻāĻŦāĻ‚ āĻŦā§āϝāĻžāĻ•āĻ—ā§āϰāĻžāωāĻ¨ā§āĻĄ āĻ•āĻ¨ā§āĻĄāĻŋāĻļāύ āϜ⧁āĻĄāĻŧ⧇ āĻŽāĻžāύ āĻĒāϰ⧀āĻ•ā§āώāĻž āĻ•āϰ⧁āύāĨ¤
  3. āĻ•āĻ–āύ āĻāĻ•āϜāύ āĻŽāĻžāύ⧁āώāϕ⧇ āφāωāϟāĻĒ⧁āϟ āĻĒāĻ°ā§āϝāĻžāϞ⧋āϚāύāĻž āĻŦāĻž āĻ…āύ⧁āĻŽā§‹āĻĻāύ āĻ•āϰāϤ⧇ āĻšāĻŦ⧇ āϤāĻž āύāĻŋāĻ°ā§āϧāĻžāϰāĻŖ āĻ•āϰ⧁āύāĨ¤
  4. āϏāĻŋāĻ¨ā§āĻĨ⧇āϟāĻŋāĻ• āĻ…āĻĄāĻŋāĻ“ āϞ⧇āĻŦ⧇āϞ āĻ•āϰ⧁āύ āĻāĻŦāĻ‚ āĻĻāĻžāϝāĻŧāĻŦāĻĻā§āϧāϤāĻžāϰ āϜāĻ¨ā§āϝ āĻŽā§‚āϞ āϰ⧇āĻ•āĻ°ā§āĻĄ āϰāĻžāϖ⧁āύāĨ¤

āϏ⧂āĻ¤ā§āϰ āĻāĻŦāĻ‚ āφāϰāĻ“ āĻĒāĻĄāĻŧ⧁āύ

  1. arXiv / IEEE MMSP 2018āϰāĻŋāϝāĻŧ⧇āϞ-āϟāĻžāχāĻŽ āĻĢ⧁āϞ-āĻŦā§āϝāĻžāĻ¨ā§āĻĄ āĻ¸ā§āĻĒāĻŋāϚ āĻāύāĻšāĻžāĻ¨ā§āϏāĻŽā§‡āĻ¨ā§āĻŸā§‡āϰ āϜāĻ¨ā§āϝ āĻāĻ•āϟāĻŋ āĻšāĻžāχāĻŦā§āϰāĻŋāĻĄ āĻĄāĻŋāĻāϏāĻĒāĻŋ/āĻĄāĻŋāĻĒ āϞāĻžāĻ°ā§āύāĻŋāĻ‚ āĻ…ā§āϝāĻžāĻĒā§āϰ⧋āϚ
  2. Jean-Marc Valin (Xiph.Org / Mozilla)RNNoise: āĻļ⧇āĻ–āĻžāϰ āĻļāĻŦā§āĻĻ āĻĻāĻŽāύ
  3. Xiph.Org Foundation (GitHub)xiph/rnnoise: āĻ…āĻĄāĻŋāĻ“ āĻļāĻŦā§āĻĻ āĻ•āĻŽāĻžāύ⧋āϰ āϜāĻ¨ā§āϝ āĻĒ⧁āύāϰāĻžāĻŦ⧃āĻ¤ā§āϤ āύāĻŋāωāϰāĻžāϞ āύ⧇āϟāĻ“āϝāĻŧāĻžāĻ°ā§āĻ•
  4. PhoronixāύāĻŋāωāϰāĻžāϞ āύ⧇āϟāĻ“āϝāĻŧāĻžāĻ°ā§āĻ• āύāϝāĻŧ⧇āϜ āϏāĻžāĻĒā§āϰ⧇āĻļāύ⧇āϰ āϜāĻ¨ā§āϝ AVX2 āĻ…āĻĒā§āϟāĻŋāĻŽāĻžāχāĻœā§‡āĻļāĻžāύ āϏāĻš RNNoise 0.2 āĻĒā§āϰāĻ•āĻžāĻļ āĻ•āϰāĻž āĻšāϝāĻŧ⧇āϛ⧇
  5. GamingOnLinuxOBS āĻ¸ā§āϟ⧁āĻĄāĻŋāĻ“ 26.0 āϰāĻŋāϞāĻŋāĻœā§‡ RNNoise-āĻāϰ āϜāĻ¨ā§āϝ āφāϰāĻ“ āĻ­āĻžāϞ āĻļāĻŦā§āĻĻ āĻĻāĻŽāύ āϝ⧋āĻ— āĻ•āϰ⧇āϛ⧇āĨ¤
  6. arXiv / INTERSPEECH 2020āĻ•āĻŽ-āϜāϟāĻŋāϞāϤāĻž, āĻĢ⧁āϞāĻŦā§āϝāĻžāĻ¨ā§āĻĄ āĻ¸ā§āĻĒāĻŋāĻšā§‡āϰ āϰāĻŋāϝāĻŧ⧇āϞ-āϟāĻžāχāĻŽ āĻŦāĻ°ā§āϧāύ⧇āϰ āϜāĻ¨ā§āϝ āĻāĻ•āϟāĻŋ āωāĻĒāϞāĻŦā§āϧāĻŋāĻŽā§‚āϞāĻ•āĻ­āĻžāĻŦ⧇-āĻĒā§āϰ⧇āϰāĻŖāĻžāĻŽā§‚āϞāĻ• āĻĒāĻĻā§āϧāϤāĻŋ
  7. arXiv / INTERSPEECH 2020INTERSPEECH 2020 āĻ—āĻ­ā§€āϰ āĻļāĻŦā§āĻĻ āĻĻāĻŽāύ āĻšā§āϝāĻžāϞ⧇āĻžā§āϜ: āĻĄā§‡āϟāĻžāϏ⧇āϟ, āĻŦāĻŋāώāϝāĻŧāĻ­āĻŋāĻ¤ā§āϤāĻŋāĻ• āĻĒāϰ⧀āĻ•ā§āώāĻžāϰ āĻĢā§āϰ⧇āĻŽāĻ“āϝāĻŧāĻžāĻ°ā§āĻ• āĻāĻŦāĻ‚ āĻšā§āϝāĻžāϞ⧇āĻžā§āĻœā§‡āϰ āĻĢāϞāĻžāĻĢāϞ

āĻ…āĻ¨ā§āĻŦ⧇āώāĻŖ āϚāĻžāϞāĻŋāϝāĻŧ⧇ āϝāĻžāύ

āĻĢā§āϰāĻŋ āύāĻŋāωāϜāϞ⧇āϟāĻžāϰ

āĻĒā§āϰāϤāĻŋāĻĻāĻŋāύ⧇āϰ āĻāφāχ āĻŦā§āϰāĻŋāĻĢāĻŋāĻ‚ āĻĒāĻžāύ

āĻĒā§āϰāϤāĻŋ āϏāĻĒā§āϤāĻžāĻšā§‡āϰ āĻ•āĻžāĻœā§‡āϰ āĻĻāĻŋāύ āϏāĻ•āĻžāϞ⧇ āϤāĻŋāύāϟāĻŋ āϝāĻžāϚāĻžāχāĻ•ā§ƒāϤ āĻāφāχ āĻ¸ā§āĻŸā§‹āϰāĻŋ, āϏāĻšāϜ āχāĻ‚āϰ⧇āϜāĻŋāϤ⧇ āϞ⧇āĻ–āĻžāĨ¤ āϚāĻŋāϰāĻ•āĻžāϞ⧇āϰ āϜāĻ¨ā§āϝ āĻĢā§āϰāĻŋ, āϕ⧋āύ⧋ āĻŦāĻŋāĻœā§āĻžāĻžāĻĒāύ āύ⧇āχāĨ¤

āĻĒā§āϰāϤāĻŋ āϏāĻĒā§āϤāĻžāĻšā§‡āϰ āĻ•āĻžāĻœā§‡āϰ āĻĻāĻŋāύ⧇ āĻāĻ•āϟāĻŋ āχāĻŽā§‡āϞāĨ¤ āĻāĻ• āĻ•ā§āϞāĻŋāϕ⧇āχ āφāύāϏāĻžāĻŦāĻ¸ā§āĻ•ā§āϰāĻžāχāĻŦ āĻ•āϰ⧁āύāĨ¤ āφāĻŽāϰāĻž āĻ•āĻ–āύ⧋āχ āφāĻĒāύāĻžāϰ āĻ āĻŋāĻ•āĻžāύāĻž āĻŦāĻŋāĻ•ā§āϰāĻŋ āĻŦāĻž āĻļā§‡ā§ŸāĻžāϰ āĻ•āϰāĻŋ āύāĻžāĨ¤

āύāĻŋāĻœā§‡āϕ⧇ āĻĒāϰ⧀āĻ•ā§āώāĻž āĻ•āϰ⧁āύ

RNNoise āϕ⧁āχāĻœā§‡āϰ āϏāĻžāĻĨ⧇ āĻ¸ā§āĻĒāĻŋāϚ āĻĄāĻŋāύ⧋āχāϏāĻŋāĻ‚ āύāĻŋāύ

āĻĒā§āϰāϤāĻŋāϟāĻŋ āωāĻ¤ā§āϤāϰ⧇āϰ āϤāĻžā§ŽāĻ•ā§āώāĻŖāĻŋāĻ• āĻĢāĻŋāĻĄāĻŦā§āϝāĻžāĻ•, āĻāĻŦāĻ‚ āϕ⧋āĻ°ā§āϏ āĻĒāĻžāϏ āĻ•āϰāĻžāϰ āĻĒāϰ āϝāĻžāϚāĻžāχāϝ⧋āĻ—ā§āϝ āφāχāĻĄāĻŋ āϏāĻš āĻāĻ•āϟāĻŋ āĻļā§‡ā§ŸāĻžāϰāϝ⧋āĻ—ā§āϝ āϏāĻžāĻ°ā§āϟāĻŋāĻĢāĻŋāϕ⧇āϟāĨ¤

āϕ⧁āχāϜ āĻļ⧁āϰ⧁ āĻ•āϰ⧁āύ

āĻĢā§āϰāĻŋ āĻāφāχ āĻļāĻŋāĻ•ā§āώāĻžāϕ⧇ āϏāĻŽāĻ°ā§āĻĨāύ āĻ•āϰ⧁āύāĨ¤ AI Understanding āĻāĻ•āϟāĻŋ 501(c)(3) āĻ…āϞāĻžāĻ­āϜāύāĻ• āĻĒā§āϰāϤāĻŋāĻˇā§āĻ āĻžāύ — āϕ⧋āύ⧋ āĻŦāĻŋāĻœā§āĻžāĻžāĻĒāύ āύ⧇āχ, āϕ⧋āύ⧋ āĻĒ⧇āĻ“ā§ŸāĻžāϞ āύ⧇āχ, āĻ•āĻ–āύ⧋āχ āύāĻžāĨ¤ āĻāĻ•āϟāĻŋ āĻĻāĻžāύ āĻ•āϰ⧁āύ

āĻĒā§āϰāĻžāϝāĻŧāĻļāχ āϜāĻŋāĻœā§āĻžāĻžāϏāĻŋāϤ āĻĒā§āϰāĻļā§āύāĻžāĻŦāϞ⧀

RNNoise āĻĻāĻŋāϝāĻŧ⧇ āĻ¸ā§āĻĒāĻŋāϚ āĻĄāĻŋāύ⧋āχāϏāĻŋāĻ‚ āĻ•āĻŋ?

RNNoise āĻšāϞ āĻāĻ•āϟāĻŋ āĻ“āĻĒ⧇āύ-āϏ⧋āĻ°ā§āϏ āύāϝāĻŧ⧇āϜ āϏāĻžāĻĒā§āϰ⧇āĻļāύ āϞāĻžāχāĻŦā§āϰ⧇āϰāĻŋ, āϝāĻž Xiph.Org āĻāĻŦāĻ‚ Mozilla-āĻāϰ āϜāĻŋāύ-āĻŽāĻžāĻ°ā§āĻ• āĻ­ā§āϝāĻžāϞāĻŋāύ āĻĻā§āĻŦāĻžāϰāĻž 2017 āϏāĻžāϞ⧇ āĻĒā§āϰāĻ•āĻžāĻļāĻŋāϤ āĻšāϝāĻŧ⧇āϛ⧇, āϝāĻž āϰāĻŋāϝāĻŧ⧇āϞ āϟāĻžāχāĻŽā§‡ āĻŦāĻ•ā§āϤ⧃āϤāĻž āĻĨ⧇āϕ⧇ āĻĒāϟāĻ­ā§‚āĻŽāĻŋāϰ āĻļāĻŦā§āĻĻ āϏāϰāĻŋāϝāĻŧ⧇ āĻĻ⧇āϝāĻŧāĨ¤ āĻāϟāĻŋ āĻĒā§āϰāĻžāϝāĻŧ 88,000 āĻ“āϜāύ⧇āϰ āĻāĻ•āϟāĻŋ āϛ⧋āϟ āĻĒ⧁āύāϰāĻžāĻŦ⧃āĻ¤ā§āϤ āύāĻŋāωāϰāĻžāϞ āύ⧇āϟāĻ“āϝāĻŧāĻžāĻ°ā§āϕ⧇āϰ āϏāĻžāĻĨ⧇ āĻ•ā§āϞāĻžāϏāĻŋāĻ• āϏāĻŋāĻ—āĻ¨ā§āϝāĻžāϞ āĻĒā§āϰāϏ⧇āϏāĻŋāĻ‚ āϝ⧁āĻ•ā§āϤ āĻ•āϰ⧇, āϤāĻžāχ āĻāϟāĻŋ āĻāĻ•āϟāĻŋ āϏāĻžāϧāĻžāϰāĻŖ CPU āϕ⧋āϰ⧇ āϚāϞ⧇ āĻāĻŦāĻ‚ āĻ¸ā§āĻŸā§āϰāĻŋāĻŽāĻŋāĻ‚ āĻāĻŦāĻ‚ āĻ­āϝāĻŧ⧇āϏ-āĻ•āϞ āϏāĻĢā§āϟāĻ“āϝāĻŧā§āϝāĻžāϰ⧇āϰ āĻāĻ•āϟāĻŋ āϏāĻžāϧāĻžāϰāĻŖ āĻŦāĻŋāĻ˛ā§āĻĄāĻŋāĻ‚ āĻŦā§āϞāĻ• āĻšāϝāĻŧ⧇ āĻ“āϠ⧇āĨ¤

RNNoise āĻĒā§āϰāĻžāĻĨāĻŽāĻŋāĻ•āĻ­āĻžāĻŦ⧇ āĻ…āĻĄāĻŋāĻ“āϰ āĻĒā§āϰāϤāĻŋāϟāĻŋ āϛ⧋āϟ āĻĢā§āϰ⧇āĻŽā§‡āϰ āϜāĻ¨ā§āϝ āϕ⧀ āĻ­āĻŦāĻŋāĻˇā§āϝāĻĻā§āĻŦāĻžāĻŖā§€ āĻ•āϰ⧇?

RNNoise āĻĒā§āϰāϤāĻŋ-āĻŦā§āϝāĻžāĻ¨ā§āĻĄā§‡āϰ āϞāĻžāϭ⧇āϰ āĻ…āύ⧁āĻŽāĻžāύ āĻ•āϰ⧇ āϝāĻž āĻĒā§āϰāϤāĻŋāϟāĻŋ āĻŦāĻ°ā§āĻŖāĻžāϞ⧀ āĻŦāĻŋāύ⧇ āĻ•āĻžāϜ āĻ•āϰāĻžāϰ āĻĒāϰāĻŋāĻŦāĻ°ā§āϤ⧇ āĻļāĻŦā§āĻĻ-āĻĒā§āϰāϧāĻžāύ āĻŦā§āϝāĻžāĻ¨ā§āĻĄāϗ⧁āϞāĻŋāϕ⧇ āϏāĻ‚āϰāĻ•ā§āώāĻŖ āĻ•āϰāĻžāϰ āϏāĻŽāϝāĻŧ āĻļāĻŦā§āĻĻ-āĻĒā§āϰāϧāĻžāύ āĻŦā§āϝāĻžāĻ¨ā§āĻĄāϗ⧁āϞāĻŋāϕ⧇ āĻšā§āϰāĻžāϏ āĻ•āϰ⧇āĨ¤

RNNoise āĻāϰ āĻŽā§‚āϞ āĻ…āĻ‚āĻļ⧇ āϕ⧋āύ āϧāϰāύ⧇āϰ āύāĻŋāωāϰāĻžāϞ āύ⧇āϟāĻ“āϝāĻŧāĻžāĻ°ā§āĻ• āϰāϝāĻŧ⧇āϛ⧇?

RNNoise āĻāĻ•āϟāĻŋ āĻ•āĻŽāĻĒā§āϝāĻžāĻ•ā§āϟ āϰāĻŋāĻ•āĻžāϰ⧇āĻ¨ā§āϟ āύāĻŋāωāϰāĻžāϞ āύ⧇āϟāĻ“āϝāĻŧāĻžāĻ°ā§āĻ• āĻŦā§āϝāĻŦāĻšāĻžāϰ āĻ•āϰ⧇ āϗ⧇āĻŸā§‡āĻĄ āϰāĻŋāĻ•āĻžāϰ⧇āĻ¨ā§āϟ āχāωāύāĻŋāĻŸā§‡āϰ āϏāĻžāĻĨ⧇ āϤ⧈āϰāĻŋ, āĻāϟāĻŋ āϛ⧋āϟ āĻĨāĻžāĻ•āĻžāϰ āϏāĻŽāϝāĻŧ āĻŸā§‡āĻŽā§āĻĒā§‹āϰāĻžāϞ āĻŽā§‡āĻŽāϰāĻŋ āĻĻ⧇āϝāĻŧāĨ¤

āϕ⧇āύ RNNoise āĻ•āĻžāρāϚāĻž āĻŦāĻ°ā§āĻŖāĻžāϞ⧀ āĻŦāĻŋāύ⧇āϰ āĻĒāϰāĻŋāĻŦāĻ°ā§āϤ⧇ āĻ…āύ⧁āϧāĻžāĻŦāύāϝ⧋āĻ—ā§āϝ āĻĢā§āϰāĻŋāϕ⧋āϝāĻŧ⧇āĻ¨ā§āϏāĻŋ āĻŦā§āϝāĻžāĻ¨ā§āĻĄ āĻŦā§āϝāĻŦāĻšāĻžāϰ āĻ•āϰ⧇?

āĻļ⧁āϧ⧁āĻŽāĻžāĻ¤ā§āϰ ~22 āĻŦā§āϝāĻžāĻ¨ā§āĻĄ āϞāĻžāϭ⧇āϰ āĻ­āĻŦāĻŋāĻˇā§āϝāĻĻā§āĻŦāĻžāĻŖā§€ āĻ•āϰāĻž āύ⧇āϟāĻ“āϝāĻŧāĻžāĻ°ā§āĻ•āϟāĻŋāϕ⧇ āϛ⧋āϟ, āĻĻā§āϰ⧁āϤ āĻāĻŦāĻ‚ āĻŦāĻžāĻĻā§āϝāϝāĻ¨ā§āĻ¤ā§āϰ-āĻļāĻŦā§āĻĻ⧇āϰ āφāĻ°ā§āϟāĻŋāĻĢā§āϝāĻžāĻ•ā§āĻŸā§‡āϰ āĻĒā§āϰāϤāĻŋ āĻ•āĻŽ āĻĒā§āϰāĻŦāĻŖ āϰāĻžāϖ⧇ āϝāĻž āĻĒā§āϰāϤāĻŋ-āĻŦāĻŋāύ āĻĒāĻĻā§āϧāϤāĻŋāϤ⧇ āĻĒā§āϞ⧇āĻ— āĻ•āϰ⧇āĨ¤

RNNoise āĻŽāĻĄā§‡āϞāϟāĻŋ āĻŽā§‹āϟāĻžāĻŽā§āϟāĻŋ āĻ•āϤ āĻŦāĻĄāĻŧ?

RNNoise-āĻāϰ āĻ…āĻ°ā§āĻĄāĻžāϰ⧇ 85,000 āĻ“āϜāύ āϰāϝāĻŧ⧇āϛ⧇, āϝāĻž āĻāĻ•āϟāĻŋ āϏāĻŋāĻĒāĻŋāχāω āϕ⧋āϰ⧇ āϰāĻŋāϝāĻŧ⧇āϞ āϟāĻžāχāĻŽā§‡āϰ āĻšā§‡āϝāĻŧ⧇ āĻĻā§āϰ⧁āϤ āϚāĻžāϞāĻžāύ⧋āϰ āϜāĻ¨ā§āϝ āϝāĻĨ⧇āĻˇā§āϟ āϛ⧋āϟāĨ¤

RNNoise-āĻ āĻĒāĻŋāϚ āĻĢāĻŋāĻ˛ā§āϟāĻžāϰ⧇āϰ āĻ­ā§‚āĻŽāĻŋāĻ•āĻž āϕ⧀?

āĻāĻ•āϟāĻŋ āϚāĻŋāϰ⧁āύāĻŋ/āĻĒāĻŋāϚ āĻĢāĻŋāĻ˛ā§āϟāĻžāϰ āĻŦā§āϝāĻžāĻ¨ā§āĻĄ āϞāĻžāϭ⧇āϰ āĻĒāϰāĻŋāĻĒā§‚āϰāĻ•, āĻšāĻžāϰāĻŽā§‹āύāĻŋāĻ•ā§āϏ⧇āϰ āĻŽāĻ§ā§āϝ⧇ āĻŦāϏ⧇ āĻĨāĻžāĻ•āĻž āĻļāĻŦā§āĻĻāϕ⧇ āĻĻāĻŽāύ āĻ•āϰāϤ⧇ āĻ•āĻŖā§āĻ āĻ¸ā§āĻŦāϰāϝ⧁āĻ•ā§āϤ āĻŦāĻ•ā§āϤ⧃āϤāĻžāϰ āϏ⧁āϰ⧇āϞāĻž āĻ•āĻžāĻ āĻžāĻŽā§‹āϕ⧇ āĻ•āĻžāĻœā§‡ āϞāĻžāĻ—āĻžāϝāĻŧāĨ¤

āĻļāĻŋāĻ–āϤ⧇ āĻĨāĻžāϕ⧁āύ

āϏāĻŽā§āĻĒāĻ°ā§āĻ•āĻŋāϤ āĻ—āĻžāχāĻĄ

āĻāχ āĻŦāĻŋāώāϝāĻŧ⧇āϰ āϜāĻ¨ā§āϝ āĻŦāĻžāĻ›āĻžāχ āĻ•āϰāĻž āφāϰāĻ“ āĻ—āĻžāχāĻĄ

āĻĒāϰāĻŦāĻ°ā§āϤ⧀ āφāĻĒāĻĒāϰāĻŦāĻ°ā§āϤ⧀ āĻ—āĻžāχāĻĄ

āĻŦāĻ•ā§āϤ⧃āϤāĻž āĻŦāĻŋāĻšā§āϛ⧇āĻĻ āĻāĻŦāĻ‚ āĻ•āĻ•āĻŸā§‡āϞ āĻĒāĻžāĻ°ā§āϟāĻŋ āϏāĻŽāĻ¸ā§āϝāĻž

āĻ…āĻĄāĻŋāĻ“ āĻāφāχ

āĻāĻ•āϟāĻŋ āĻŦāĻŋāύāĻžāĻŽā§‚āĻ˛ā§āϝ⧇ AI āĻļāĻ‚āϏāĻžāĻĒāĻ¤ā§āϰ āĻĒāĻžāύāĻāĻ•āϟāĻŋ āϏāĻ‚āĻ•ā§āώāĻŋāĻĒā§āϤ āĻŦāĻŋāύāĻžāĻŽā§‚āĻ˛ā§āϝ⧇ āϕ⧋āĻ°ā§āϏ āĻāĻŦāĻ‚ āĻĒāϰ⧀āĻ•ā§āώāĻž. āϕ⧋āύ⧋ āĻ–āϰāϚ āύ⧇āχ, āϕ⧋āύ⧋ āĻ•āĻžāĻ°ā§āĻĄ āύ⧇āχāĨ¤ āĻĒā§āϰāĻžāϝāĻŧ 2.5 āϘāĻ¨ā§āϟāĻžāĨ¤āĻŦāĻŋāύāĻžāĻŽā§‚āĻ˛ā§āϝ⧇ āĻļ⧁āϰ⧁ āĻ•āϰ⧁āύ →