faster-whisper large-v3
Free.ai (self-hosted)
·
stt
·
~500 tohu i ia minute
Ko te faster-whisper large-v3 he a tauira-whakaahua-ki-te-tuhi i hangaia e te OpenAI / SYSTRAN. He kaha rawa i te Accurate transcription. I whakanohoia ki ngā GPUs Free.ai - e haere ana i te wātea i runga i tō tātau pūpū tohu i ia rā (500 ngā tohu i ia wa). I tukua i raro i te MIT — i whakaaetia te whakamahinga hokohoko i runga i te Free.ai.
Ka whakamahia mā te API
API REST OpenAI-hōia. Whakana tētahi kī, ā, ka karangatia tēnei tauira i roto i ngā takirua.
curl -X POST https://api.free.ai/v1/stt/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"faster-whisper-large-v3","audio_url":"https://..."}'He ōrite nga tauira
E pā ana ngā pātai
Ka whakamāoritia e faster-whisper large-v3 te oro i kōrerotia ki roto i te kupu. Whakarewa i tētahi MP3, WAV, M4A, he faila ataata rānei, ā, ka hoki mai a faster-whisper large-v3 ki te whakamāoritanga katoa me ngā whakarārangi SRT/VTT kōwhiria me ngā tātaitai wā.
He tokomaha nga reo e whakahaeretia ana e faster-whisper large-v3 — ko ngā tauira whānau Whisper e taupoki ana i te 90%, Parakeet e taupoki ana i te ~25, e rerekē ana ētahi atu. Ka tīpako "mātau-mātau", ka whakapūtā rānei i te reo mō te tika tiketike rawa.
Ko te ōrautanga hē-wānanga-wānanga ko te 5–10% i runga i te oro Ingarihi mārama, 10–20% i runga i te oro whakahauhau, i te oro whakahauhau rānei.
He — kei roto i ia wāhanga ngā tātai tīmata/whakamutu. Ka whakaputaina hei SRT, VTT rānei me te mahere wā ki runga i tōna ataata.
E haere ana a faster-whisper large-v3 ki a tātau ake GPUs ki tōtou motika wātea i te tuatahi; $5 → 200,000 ngā tohu whakawhiwhita i muri mai. Neke atu i ngā tohu ~500 i ia minu.
MP3, WAV, M4A, FLAC, OGG, tae atu ki te ataata (MP4, MOV, WebM) — ka tangohia e tātau te oro. 500 MB te nui rawa o ia tāpiritanga. He nui ake ngā faila? Whakawehe me /oro/ whakaiti/, hoatu rānei i /v1/stt/batch/.
He whakawhitinga motuhake te whakahua i te kaikōrero — ka whakawhiti "pāpāho" i runga i te / whakamāori /. faster-whisper large-v3 e whakahaere ana i te whakamāoritanga; ko te whakahua i ngā tohu i ia wāhanga me te Kaikōrero 1 / Kaikōrero 2 / ērā atu mea.
He — /batch/ e whakaae ana ki tētahi pūrākau o ngā pūranga oro. Ka tae ia whakahua i roto i te /account/?tab=hitori me te ingoa pūranga taketake. Mō te pupuritanga o te rākau pūrākau, ka whakamahia te API.
He — POST tōtou oro ki te /v1/stt/transcribe/ me te tauira "faster-whisper large-v3". Ka hoki mai te JSON me te kupu + wāhanga + wā-wā. He tohutoro katoa te /api/.
Ko ngā tauira ā-whāinga e pupuri ana i te oro i runga i a tātau GPU; ka whakawhitia te utu mā te DPA. Ka tangohia te oro i muri i te matapihi tiritiri (24h anon, 7d te whakaingoatanga). Kāore e whakaakona e tātau i ōna tāurunga.
He — Free.ai e whakaae ana ki te whakamahi hokohoko o ngā tāruatanga. E hiahiatia ana e koe ngā mana ki te oro i whakarewaina e koe (ko tō koe te whakataki, ngā rawa whakawhiwhinga, ngā ihirangi rānei me te whakaaetanga).
Ko te take o te wā tūturu tata tonu ki te 0.05–0.2× — he podcast 60-meneti e whakarerekē ana i roto i ngā minu 3–12. He tere ake te mutunga o ngā tauira Premium. Ka whakamahia te kī kōaro hei kati i te tīpako.