{"name":"Whisper","tagline":"OpenAI's open-weights speech recognition — the default transcription model.","url":"https://github.com/openai/whisper","category":"audio","pricing":"free","price_note":"open source (MIT), self-host free; hosted: gpt-4o-transcribe $0.006/min","free_tier":true,"open_source":true,"api":true,"self_host":true,"models_used":["whisper-v3"],"verdict":"ship","verdict_text":"The main reason paid transcription is optional now. large-v3 is accurate across dozens of languages and runs on a laptop. It still hallucinates on silence and noisy segments, doesn't do speaker diarization out of the box, and needs a decent GPU for real-time. Use it as the baseline for any transcription pipeline — hosted or self-hosted.","limitations":["Hallucinates text on silence and noisy segments.","No built-in speaker diarization (pair with pyannote).","large-v3 needs a decent GPU for real-time work.","whisper-1 API is legacy; the current hosted line is the gpt-transcribe family."],"strengths":["The reason paid transcription is optional: large-v3 is accurate across dozens of languages.","Open weights that run on a laptop; hosted gpt-transcribe exists when you do not want ops."],"use_for":["A baseline STT pipeline, hosted or self-hosted.","Offline or private transcription where a vendor STT is the wrong boundary."],"skip_when":["Silence and noise would hallucinate text you cannot filter.","You need speaker diarization without pairing pyannote.","Real-time large-v3 would run without a decent GPU."],"receipts":["https://github.com/openai/whisper","https://developers.openai.com/api/docs/pricing"],"affiliate":"none","evidence_tier":"source-verified","momentum":"steady","featured":false,"data_sensitivity":{"trains_on_inputs":"unknown","local_processing":true},"last_verified":"2026-08-19","observed_by":"whysanesanders","slug":"whisper","canonical_url":"https://noemium.com/tools/whisper/","attribution":"Noemium catalog data is licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). See https://noemium.com/method/ for how entries are verified."}