Audio · by OpenAI
Whisper
OpenAI's open-source speech-recognition model that transcribes and translates audio in dozens of languages, locally or via API.
+ Free, runs fully offline, and remains the accuracy benchmark that most commercial transcription services are measured against.
− No speaker diarization out of the box, it hallucinates text during silence, and large models are slow without a GPU.
Open Whisper ↗ best for: free local transcription
Evidence
rank #2 of 257Head to head: Whisper vs AssemblyAI Whisper vs Deepgram
Whisper alternatives
All 7 alternatives →
+ Strong accuracy plus built-in audio intelligence (speaker labels, chapters, PII redaction) and LeMUR for asking questions of a transcript.
− No consumer app, and per-hour pricing adds up on high volumes compared with self-hosted Whisper.
+ Among the fastest real-time transcription APIs available, with diarization and keyword boosting built in.
− Developer-only; there is no end-user app, and accuracy on heavily accented speech trails Whisper-class models.
+ Deep integration list (Salesforce, HubSpot, Slack, Notion) and searchable transcripts across every past meeting.
− Speaker labeling and accuracy drop on noisy or accented audio, and the bot joining every calendar event by default surprises people.
+ Live transcript you can read during the call, with a generous free tier for individuals.
− The OtterPilot bot auto-joins meetings and emails attendees unless you turn it off, which annoys other participants.
+ The option to upgrade any file to a human transcript gives you a guaranteed-accuracy path for legal or published work.
− AI transcripts cost more per minute than pure-API rivals, and human turnaround can take hours to days.