WisprNote: Offline Voice Transcription
Transcribing voice notes offline with high quality is a challenge, often plagued by delays and privacy concerns. I created WisprNote, a Mac app that uses Silero VAD, FFmpeg, and OpenAI Whisper models for fast, private transcription. In the web’s silent grove, WisprNote is a whispering tree, capturing spoken thoughts with serene precision.
Silero VAD removes silence, FFmpeg processes audio, and Whisper transcribes, all offline to ensure privacy, delivering 50–200 words per second.
# Bash script snippet for WisprNote transcription
speech_file="${output_dir}/output.wav"
"$tvad" "$wav_file" "$vad_model" # Silero VAD to remove silence
"$ffmpeg" -y -i "$speech_file" -acodec pcm_s16le -ar 16000 -ac 1 "$wav_file" # Re-encode for Whisper
"$whisper" -m "$model" -otxt -of "$wav_file" "$speech_file" # Transcribe with Whisper
This script removes silence, processes audio, and transcribes with Whisper, ensuring accurate, private results.
Building WisprNote was like planting a tree in a quiet grove, nurturing thoughts into text. At DOSAYGO, we value privacy and precision, crafting tools that capture ideas securely. WisprNote is a keeper of whispers, preserving voices in a silent embrace.