NVIDIA NeMo, an open source conversational AI toolkit, announced the Parakeet ASR model family, a series of state-of-the-art automatic speech recognition (ASR) models capable of transcribing spoken English with outstanding accuracy, as reported by Webmaster Home on January 8. Nvidia announced four Parakeet models that are based on the RNN Transducer/Connectionist Temporal Classification decoder and have 0.6-110 million parameters. They are able to handle a wide range of audio environments and, after training on only 64,000 hours of datasets, achieve excellent word error rate (WER) performance on the Benchmark dataset, outperforming previous models. According to the developers, the models are robust to non-speech segments such as music and mute, and outperform OpenAI's Whisper v3 in benchmark testing. They also provide user-friendly integration into the project with pre-trained control points.