May 2019
Intermediate to advanced
272 pages
7h 19m
English
GANSynth was developed by Google's Magenta group. GANSynth is a model for audio synthesis using GANs that are trained on the NSynth dataset.
Similar to WaveGlow, GANSynth can generate audio considerably faster than its autoregressive counterparts because it can generate an entire sequence in parallel.
Another interesting feature of GANSynth is that it enables interpolation in timbre space.
For more information about GANSynth, you can visit the web page at https://storage.googleapis.com/magentadata/papers/gansynth/index.html. Alternatively, you can find the code for GANSynth at https://github.com/tensorflow/magenta/tree/master/magenta/models/gansynth.
Read now
Unlock full access