Fritter Nemotron 0.6B
No description provided
About This Project
Fritter-Nemotron-0.6B is an automatic speech recognition (ASR) model specifically fine-tuned for high-accuracy dictation of technical vocabulary, including specialized AI terminology. Built upon the NVIDIA Nemotron-speech-streaming-en-0.6b architecture, this project utilizes approximately 17 hours of targeted audio data to improve transcription performance for technical users. It is designed to support real-time streaming workflows and is distributed in multiple formats, including NeMo weights, GGUF, and sherpa-onnx, to ensure compatibility with various inference engines.
The model is particularly suited for developers and technical professionals who require reliable voice-to-text transcription for complex subject matter. By leveraging a custom dataset of read-aloud technical sentences, the model addresses common shortcomings in general-purpose ASR systems regarding specialized jargon. It is released under the CC-BY-NC-4.0 license, providing a resource for non-commercial research and personal productivity projects.