https://xiphmont.dreamwidth.org/ xiphmont - Dreamwidth Studios Fri, 29 Mar 2019 19:55:00 GMT LiveJournal / Dreamwidth Studios xiphmont personal https://v2.dreamwidth.org/11234939/2976728 https://xiphmont.dreamwidth.org/ 100 100 https://xiphmont.dreamwidth.org/95505.html Fri, 29 Mar 2019 19:55:00 GMT https://xiphmont.dreamwidth.org/95505.html <p>Jean-Marc Valin has been applying deep learning frameworks to audio over the past few years. So far he's released <a href="https://people.xiph.org/~jm/demo/rnnoise/">RNNoise</a> (a surprisingly good/fast denoising system) and <a href="https://people.xiph.org/~jm/demo/lpcnet/">LPCNet</a> (a speech synthesis system along the lines of <a href="https://arxiv.org/pdf/1609.03499.pdf">WaveNet</a>, but fast enough to use realtime on commodity hardware). <p><a href="https://people.xiph.org/~jm/demo/lpcnet_codec/">Now he's built a codec out of LPCNet.</a> <p><img src="https://people.xiph.org/~xiphmont/demo/lpcnet-codec-banner.jpg"> <p><a href="https://jmvalin.ca/papers/lpcnet_codec.pdf">"A Real-Time Wideband Neural Vocoder at 1.6 kb/s Using LPCNet"</a> presents a new wideband speech codec built out of the best parts of a brutally speed and space efficient vocoder paired with deep-learning analysis and excitation. It's alpha-grade research in a lot of ways, but decidedly not vapourware. You can <a href="https://github.com/mozilla/LPCNet/">download the source</a> and play with it now, but first, go have a look at the <a href="https://people.xiph.org/~jm/demo/lpcnet_codec/">demo page</a>.</p></p></p></p><br /><br /><img src="https://www.dreamwidth.org/tools/commentcount?user=xiphmont&ditemid=95505" width="30" height="12" alt="comment count unavailable" style="vertical-align: middle;"/> comments https://xiphmont.dreamwidth.org/95505.html xiph lpcnet jm lpc codec public 0