Apr 2, 04:07 PM

DataOcean AI Launches Dolphin ASR Model with 21.2 Million Hours of Data Supporting 40 Eastern Languages

DataOcean AI, in collaboration with Tsinghua University, has launched Dolphin, an advanced open-source Automatic Speech Recognition (ASR) model. Dolphin supports 40 Eastern languages and 22 Chinese dialects, featuring 21.2 million hours of data, of which 7.4 million hours are open data. The model is released under the Apache 2.0 license, aiming to enhance multilingual speech recognition capabilities. Additionally, ByteDanceOSS has introduced MegaTTS3, an open Text-to-Speech (TTS) model that supports English and Chinese, offering high-quality voice cloning and accent intensity control. Meanwhile, Roblox has launched a voice safety classifier that now supports seven new languages and has achieved over 23,000 downloads on GitHub and Hugging Face.

#DataOcean AI #Tsinghua University #Dolphin #Automatic Speech Recognition #Eastern #Chinese #Apache #ByteDanceOSS #MegaTTS3 #English #Roblox #GitHub #Hugging Face

Written with ChatGPT (GPT-4o mini).

Sources

Additional media

Image #1 for story dataocean-ai-launches-dolphin-asr-model-21-2-million-hours-data-supporting-40-49c5d5d6

DataOcean AI Launches Dolphin ASR Model with 21.2 Million Hours of Data Supporting 40 Eastern Languages

Sources

Additional media

Similar Stories