1DOLLY 72PACKAGE speech-to-text3REQUIRES HOST runtime@04REQUIRES HOST microphone@05REQUIRES HOST threads@067# speech-to-text, linked against the transcribe package, with the model it8# reads. Silero VAD tells speech from other sound.9FROM /v0.1.4/Dollyfile-system 915c71d4ce772989e3076060967824e48ccd6a1a6259624fe6213542df088dbc10INSTALL /v0.1.4/demos/speech/Dollyfile-parakeet 45ac10e361a0746ba603aef41eedb00a06bd70c2a1cc76894e4dbda68995e7f511USE /v0.1.4/demos/speech/Dollyfile-transcribe e097b1acdd0fe16412f3f470fa6e81d7368198131cd7f8ec7637d0b65ae5aab912USE /v0.1.4/Dollyfile-libdolly-microphone bb9377fc91ed59526ca33926fbe826999373d4c75b30acfda08704d5eae360f713REQUIRES TOOL cc14REQUIRES TOOL c++15REQUIRES TOOL tar16REQUIRES HEADER transcribe17REQUIRES LIB transcribe18SOURCE /v0.1.4/dist/static/speech/programs.tar 7482e5dc8e2a698dbc776f61ccd71b1a3a1888290ccdf94771add48c24aed96f /tmp/speech/programs.tar19SLOP tar -xf /tmp/speech/programs.tar -C /20SLOP cc -pthread -msimd128 -O2 -Wall -Werror -c /tmp/speech/speech-to-text.c -o /tmp/speech/speech-to-text.o21SLOP c++ -pthread -msimd128 -O2 /tmp/speech/speech-to-text.o /usr/lib/transcribe/libtranscribe.a \22  /usr/lib/transcribe/libggml.a /usr/lib/transcribe/libggml-cpu.a /usr/lib/transcribe/libggml-base.a -lm -o /usr/bin/speech-to-text2324SOURCE /v0.1.4/dist/static/speech/silero-vad.gguf 0d871a800c420fdd5d1b17a3aea4296d1c39608f0ff4a9295eb76457eb1c88e6 /usr/share/dolly/speech/silero-vad.gguf25EXPORTS TOOL   speech-to-text26EXPORTS FILE   silero-vad          /usr/share/dolly/speech/silero-vad.gguf27EXPORTS FOLDER transcribe-licenses /usr/share/licenses/transcribe.cpp28FILE /usr/share/man/cat1/speech-to-text.129    speech-to-text: writes what the microphone hears as text, in English:30    transcribe.cpp with NVIDIA's Parakeet model and Silero's voice detector.31    32      speech-to-text [MODEL.gguf [DETECTOR.gguf]]33    34    Words appear while they are spoken: the line is shown dim and revised until35    1.5 s without speech end it; after 15 s a breath ends it, at 30 s it ends36    anyway. Ctrl+C leaves. It needs microphone@0, and the browser asks for the37    microphone, which is open only while it listens. Nothing that is said leaves38    the page: the model runs in the session.39