À suivreGuide suivant
Distil-Whisper and ASR Model Distillation
IA audio
GUIDE de l'IA audio
Whisper is an open-source automatic speech-recognition model released by OpenAI in 2022.
Écoutez ce guideNarré par l'IA · ~2 min
Its published tasks include multilingual transcription and translation into English; reported performance varies across languages and audio conditions.
The Whisper paper reports training on roughly 680,000 hours of web-collected audio with weak supervision. The model supports speech recognition, English speech translation, language identification and timestamps. The paper reports uneven language coverage and evaluates several tasks; its benchmark results are not a universal accuracy guarantee. The model card estimates the corpus as about 65% English audio with English transcripts, 18% non-English audio paired with English translation text and 17% non-English audio with original-language transcripts.
Hypothetical workflow: transcribe a recorded interview, then check proper nouns and uncertain spans against the audio.
Hypothetical workflow: translate non-English speech into English text and verify the translation with a qualified speaker when accuracy matters.
Use language-identification output as a model prediction that may be less reliable for underrepresented languages or noisy audio.
For long recordings, inspect segment boundaries and silent sections for omissions, repetition loops or hallucinated text.
Use the published paper and model card to select evaluation data that matches the intended language, audio quality and task. Report per-language and task-specific error rates and manually verify high-impact transcripts. The cited sources do not establish a current adoption trend or guarantee future on-device/streaming capabilities.
Newsletter gratuite
Trois histoires d'IA vérifiées chaque matin de la semaine, écrites dans un anglais simple. Gratuit pour toujours, sans publicité.
Un e-mail chaque jour de la semaine. Désabonnez-vous en un clic. Nous ne vendons ni ne partageons jamais votre adresse.
Testez-vous
Commentaires instantanés sur chaque réponse et un certificat partageable avec une pièce d'identité vérifiable une fois que vous avez réussi un cours.
Soutenez l’enseignement gratuit de l’IA. AI Understanding est une organisation à but non lucratif 501(c)(3) : pas de publicité, pas de paywall, jamais. Faire un don
Whisper is an open-source automatic speech-recognition model released by OpenAI in 2022. Its published tasks include multilingual transcription and translation into English; reported performance varies across languages and audio conditions.
Whisper est un système de reconnaissance vocale automatique qui transcrit et traduit l'audio parlé dans de nombreuses langues.
The paper notes that the training corpus was English-heavy and that most languages had far less than 1,000 hours.
Special tokens identify the language and task, such as transcription or English translation.
Whisper gère la transcription, la traduction vers l'anglais, la détection de langue et l'horodatage, mais ce n'est pas un générateur vidéo.
The paper describes errors such as repetition loops and missing first or last words around segment boundaries.
Continuez à apprendre
Plus de guides sélectionnés pour ce sujet
À suivreGuide suivant
Distil-Whisper and ASR Model Distillation
IA audio