Voice Verter Using Whisper Algorithm

Lots of changes and innovations have come in Speech-to-Text (STT) and Text-to-Speech (TTS) technologies. STT converts speech to text, while TTS synthesizes speech from written text. The Whisper model aims to convert real-time speech in written format, summarize with Natural Language Grammar (NLG) and Meaning Representation (AMR), then convert the summary back to speech. Leveraging Deep Speech 2 and AMR graphs, Whisper approach promises efficiency. Deep Speech 2 yields a 3.2x CPU speedup whereas GPUs offer a $21 x$ acceleration for deep learning. The proposed summarizer achieves a $93.87 \%$ accuracy, better than ROGUE score of the Lead-3-AMR-Baseline model, ensuring high-quality summaries.