Distil-Whisper large-v3
Free.ai (self-hosted)
·
stt
·
~500 tokens per minute
Distil-Whisper large-v3 waa a qaabka hadalka-u-qoraalka dhisay by HuggingFace. ugu xoog badan ee Real-time transcription, large-volume batch STT.. Is-hoosaysiinta Free.ai GPUs - bilaash ayaa ka socda ishaada maalinlaha ah (500 tokens daqiiqo kasta). Soo saaray hoos MIT - isticmaalka ganacsi ee oggolaan Free.ai.
Ku isticmaal API
OpenAI-ku habboon REST API. abuuro furaha iyo wicidda noocan ah daqiiqado.
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://..."}'Su'aalaha badanaa la isweydiiyo
Distil-Whisper large-v3 transscribes maqalka hadalka in qoraalka. Upload MP3, WAV, M4A, ama faylka video iyo Distil-Whisper large-v3 ku soo laabtay qoraalka buuxa iyo sidoo kale doorasho SRT / VTT subtitles la timestamps.
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.
Word-dhibka heerka waa 5-10% on English audio nadiif ah, 10-20% on audio maqal ama accented. noocyada waaweyn ee dhismaha isku mid ah sameeyaan macno wanaagsan oo ku saabsan xaaladaha adag - dooro weyn marka audio waa adag.
Haa - qayb kasta oo ka mid ah bilow / dhamaadka timestamps. dhoofinta sida SRT ama VTT iyo wakhtiyada map si toos ah u ku saabsan video.
Distil-Whisper large-v3 socda on GPUs our gaar ah ka dhanka ah maalin kasta free pool hore; $ 5 → 200,000 ka dibna la siiyay calaamadaha. About ~ 500 calaamadaha daqiiqada.
MP3, WAV, M4A, FLAC, OGG, iyo sidoo kale video (MP4, MOV, WebM) - waxaan soo saarno audio. Max 500 MB per upload. Faylasha dheer? kala qaybsan /audio/cut/ ama isticmaal /v1/stt/batch/.
Afhayeenka diarization waa ka qayb qaataan kala duwan - toggle "diarize" on / transcribe /. Distil-Whisper large-v3 xakamaynaysaa qoraalka; diarization labels qayb kasta oo la Afhayeenka 1 / Afhayeenka 2 / iwm
Haa - / batch / aqbalaa folder files audio. transscript kastaa wuxuu ku dhacayaa / account /? tab = taariikhda leh magac file asalka ah. Si aad u folder-geed ilaalinta isticmaal API.
Haa - POST aad audio in /v1/stt/transcribe / la model="Distil-Whisper large-v3". JSON la soo celin doonaa qoraalka + qaybaha + timestamps eray-tayada. /api/ waxaa ka mid ah soo jeedinta buuxda.
Self-hosted models haystaa audio on our GPUs; premium ka gudbaan iyada oo DPA. Audio waa la tirtiri doonaa ka dib share-window (24h anon, 7d signed-in). Waxaanan tababarka ku siin in aad inputs.
Haa - Free.ai siin adeegsiga ganacsi ee qoraalada. Waxaad u baahan tahay xuquuqda audio aad soo dejisan (recording aad u gaar ah, qalabka liisan, ama content la oggolaaday).
Real-time faa'iido waa in ka badan 0.05-0.2 × - 60-daqiiqo podcast ku qorto 3-12 daqiiqo. Premium qaababka badanaa dhamaystiraan si deg deg ah.