Papers of the day   All papers

HIGH FIDELITY SPEECH SYNTHESIS WITH ADVERSARIAL NETWORKS

Comments

Sep 26 2019 DeepMind

We've developed a new model for text-to-speech using GANs (TTS-GAN), combining high quality with efficient generation. More details in the paper: https://arxiv.org/abs/1909.11646, and the abstract as read by TTS-GAN: https://drive.google.com/open?id=1yzNl7AC1IcY0Zfpi1Et6a4dsuSy-44wa
6 replies, 780 likes


Sep 26 2019 DeepMind

We've developed a new model for text-to-speech using GANs (TTS-GAN), combining high quality with efficient generation. More details in the paper: https://arxiv.org/abs/1909.11646, and the abstract as read by TTS-GAN: https://storage.googleapis.com/deepmind-media/research/abstract.wav
2 replies, 359 likes


Oct 28 2019 Gary Wang

It seems multi-resolution descriminators is all we needed to get Gans to work well for audio. Gan TTS(https://arxiv.org/abs/1909.11646) MelGan (https://arxiv.org/abs/1910.06711) WaveGan (https://arxiv.org/abs/1910.11480) Voice conversion 👀
1 replies, 50 likes


Sep 27 2019 Shawn Presser

Google used GANs to generate audio! It looks like there are some differences vs what I've been doing with StyleGAN spectrograms. From skimming the paper, it sounds like they model the waveform directly (no spectrogram).
1 replies, 7 likes


Oct 02 2019 HubBucket | Healthcare and Medicine Technology

#HealthIT #mHealth #HealthTech #NLP/#NLProc #NLU #NMT #DeepLearning ⚕️ GAN-TTS, a Generative Adversarial #NeuralNetwork for Text-to-Speech High Fidelity Speech Synthesis with Adversarial #NeuralNetworks 🖥️https://arxiv.org/abs/1909.11646?utm_source=Deep+Learning+Weekly&utm_campaign=06d703a367-EMAIL_CAMPAIGN_2019_04_24_03_18_COPY_01&utm_medium=email&utm_term=0_384567b42d-06d703a367-72970241 @HubBucket @HubIoMT @HubMobileApps @HubVoiceNLP https://t.co/HnmQEAaSTH
0 replies, 4 likes


Sep 26 2019 Mark

Please let applying the paper to the abstract become a new meme in AI research. This is nice but the summariser was cooler :-)
0 replies, 1 likes


Content