Distil-Whisper large-v3
Free.ai (self-hosted)
·
stt
·
~500 tokens per minute
Distil-Whisper large-v3 iku a Model swara-ka-teks kang dibangun déning HuggingFace. Strongest at Real-time transcription, large-volume batch STT.. Free.ai GPUs — dioperasikeun gratis ngalawan pool token anjeun saben dinten (500 tokens per minute). Dirilis ing MIT - panggunaan komersial diijini ing Free.ai.
Nggunakake API
API REST OpenAI-kompatibel. Nyiptakeun kunci sarta ngajual model ieu dina detik.
curl -X POST https://api.free.ai/v1/stt/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"distil-whisper-large-v3","audio_url":"https://..."}'Takon-takon sing asring diajukake
Distil-Whisper large-v3 ngatranskripsikeun audio anu diucapkeun kana teks. Ngauploadkeun MP3, WAV, M4A, atawa file video sarta Distil-Whisper large-v3 bakal ngabalikeun transkripsi pinuh plus subtitle SRT/VTT pilihan kalawan timestamp.
Distil-Whisper large-v3 handles dozens of languages - Whisper-family models cover 90+, others vary. Pick "auto-detect" or specify the language for highest accuracy.
Tingkat kasalahan kecap nyaéta 5-10% dina audio Inggris anu bersih, 10-20% dina audio anu bising atanapi diaksés. Variasi ageung tina arsitektur anu sami langkung saé dina kasus anu hadé - pilih anu langkung ageung nalika audiona kasar.
Ya - saben segmen ngandung tanda wektu wiwitan/akhiran. Eksport minangka SRT utawa VTT lan peta wektu langsung ka video anjeun.
Distil-Whisper large-v3 dijalankeun dina GPUs urang sorangan ngalawan kolam renang bébas anjeun sapopoé mimiti; $5 → 200,000 tokens dibayar sanggeus éta. Sabudeureun ~ 500 tokens per menit.
MP3, WAV, M4A, FLAC, OGG, plus video (MP4, MOV, WebM) - kita ngekstrak audio. Max 500 MB saben upload. Files luwih dawa? Dipisahake karo /audio/cut/ utawa nganggo /v1/stt/batch/.
Diaresis pangucapan (speaker diarization) iku pasa sing béda - toggle "diarize" on /transcribe/. Distil-Whisper large-v3 ngawengku transkripsi; diaresis label saben segmen karo Speaker 1 / Speaker 2 / etc.
Ya - /batch/ nampi folder berkas audio. Satiap transcript disimpen dina /account/?tab=history kalayan nami berkas aslina. Pikeun ngajaga pohon folder, gunakeun API.
Ya - POST audiomu menyang /v1/stt/transcribe/ karo model="Distil-Whisper large-v3". Ngulang JSON kanthi teks + segmen + timestamp tingkat tembung. /api/ duwé referensi lengkap.
Model self-hosted nyimpen audio dina GPU urang; premium ngaliwatan DPA. Audio dipaling saatos jendela babagi (24 jam anonim, 7 dinten ngadaptar). Kami henteu diajarkeun kana input anjeun.
Ya - Free.ai ngayakeun pamakéan komersial tina transcript. Anjeun butuh hak kana audio anu anjeun unggah (rekaman anjeun sorangan, bahan anu dilisensikan, atanapi isi anu disetujui).
Faktor real-time kira-kira 0.05-0.2× - a 60-minute podcast transcribes in 3-12 minutes. Premium models often finish faster. Use the queue button to close the tab.