Insanely Fast Whisper is a tool powered by transformers, optimum, and flash-attn that allows users to transcribe 300 minutes of audio in less than 98 seconds. The tool offers several optimization types for transcription speed and accuracy, including Transformers (fp32), Distil-Whisper (fp16 + batching [24] + bettertransformer), Faster Whisper (8-bit + beam_size [1]), among others. Users can access the tool via a command-line interface or through code using torch and the transformers pipeline.