IBM updates Granite Speech with models for translation, speed and speaker tracking

IBM's speech strategy lead George Saon and a team of IBM researchers updated Granite Speech on October 21 with three models: a new 2B multilingual transcription and translation model plus revisions to two existing models for high-throughput speech recognition and speaker- and timing-tracking transcripts. IBM reported the speed and accuracy figures, which the company says give buyers a starting point but still require testing on their own audio.

IBM updates Granite Speech with models for translation, speed and speaker tracking George Saon and IBM's speech team added a 2B multilingual model and revised two others, with performance figures largely reported by IBM. By RuntimeWire Staff https://runtimewire.com/author/runtimewire-staff · Published Primary source: Hugging Face Newsroom https://huggingface.co/blog/ibm-granite/granite-speech-5 Why it matters IBM is packaging speech AI around distinct deployment constraints, with separate models for multilingual translation, high-throughput transcription and speaker tracking. Its reported speed and accuracy results give buyers a starting point, but they still need to test the models on their own audio. George Saon https://research.ibm.com/people/george-saon?ref=runtimewire , IBM's https://www.ibm.com/?ref=runtimewire speech strategy lead, and a team of researchers updated Granite Speech https://www.ibm.com/granite?ref=runtimewire on October 21st with three models aimed at different jobs: multilingual transcription and translation, high-throughput speech recognition, and transcripts that track speakers and timing. The Hugging Face announcement https://huggingface.co/blog/ibm-granite/granite-speech-5?ref=runtimewire adds one new model and revises two existing ones.… When you purchase through links in our articles, we may earn a small commission https://runtimewire.com/affiliate-monetization-standards . This doesn’t affect our editorial independence.