whisper

Transcribe audio files with timestamps by processing spoken audio through speech-to-text models to generate segment or word-level timing for subtitle workflows. The transcription supports multilingual recordings and outputs accurate text with precise timing.

Multilingual transcription supports 99 languages with automatic language detection, translating non-English speech into English text. This handles multilingual recordings and identifies the spoken language automatically without manual configuration.

Speech-to-text processing handles noisy audio and domain-specific recordings by using flexible model sizes from tiny to large. Optional language prompts and GPU acceleration improve transcription robustness across challenging audio conditions and domain vocabulary.

Generate meeting notes and podcast transcripts by processing audio files through speech-to-text models that convert spoken content into accurate text. Batch processing handles multiple audio files for efficient podcast transcription and meeting note generation workflows.

ffmpeg is required for audio input processing to handle various audio formats before speech-to-text conversion. The transcription workflow uses ffmpeg audio input with OpenAI Whisper models ranging from tiny to large for flexible processing.