whisper

OpenAI's general-purpose speech recognition model supporting 99 languages, transcription, translation to English, and language identification, with six model sizes from tiny (39M parameters) to large (1550M parameters).

Tasks

Transcribe podcast or meeting audio to text with timestamps using the whisper CLI or Python API
Translate non-English audio into English text using task='translate'
Generate SRT/VTT subtitle files from video audio
Batch transcribe multiple audio files and save text outputs

Inputs

Audio files such as MP3, WAV (WAV recommended for better compatibility; audio can be extracted from video with ffmpeg)
Optional transcription options: language, task (transcribe/translate), initial_prompt, word_timestamps, temperature

Outputs

Transcription text with segment-level start/end timestamps (plain text, SRT, VTT, or JSON via CLI)
English-translated text from non-English audio
Word-level timestamps when word_timestamps=True

Limitations and checks

May hallucinate or invent repeated text
Accuracy degrades on audio longer than 30 minutes
No speaker diarization
Quality varies by accent and language
Background noise can affect accuracy
Not suitable for live captioning due to real-time latency
Confirm ffmpeg is installed before transcription (brew install ffmpeg on macOS, apt install ffmpeg on Ubuntu, choco install ffmpeg on Windows)
Verify Python version is 3.8-3.11 as required by the install instructions

AI Search

Find projects, verify facts, compare options, or turn a complex need into an actionable plan

Try a searchA click only fills the search box; you stay in control

Collection Scenario Project type License Deployment Resource form

whisper

Tasks

Inputs

Outputs

Limitations and checks

AI Search

Project Details