Moonshine Base
Free.ai (self-hosted)
·
stt
·
~500 Ii-token minute
Moonshine Base yi an imodeli yokuthetha-uku-kubhala eyenziwe ngu Useful Sensors. Enamandla kakhulu kwi Low-latency live transcription, embedded devices.. I-self-hosted kwi-Free.ai GPUs — iqhuba ngokukhululekileyo kwi-token pool yakho yosuku (500 tokens Iinketho ze projekti). Ikhutshwe phantsi kwe MIT - ukusetyenziswa korhwebo kuvunyelwe kwi Free.ai.
Sebenzisa nge API
I-REST API ehambelana ne-OpenAI. Yenza iqhosha kwaye unxulumane nale modeli kwimizuzu.
curl -X POST https://api.free.ai/v1/stt/ \
-H "Authorization: Bearer sk-free-..." \
-H "Content-Type: application/json" \
-d '{"model":"moonshine-base","audio_url":"https://..."}'Imibuzo ebuzwa rhoqo
Moonshine Base iguqula umbhalo othethayo ube ngumbhalo. Layisha phezulu i MP3, WAV, M4A, okanye ifayili yevidiyo kwaye Moonshine Base ibuyisela uguqulo olupheleleyo kunye ne-SRT/VTT ekhethiweyo ebhalwe phantsi nge-timestamp.
Moonshine Base handles dozens of languages - Whisper-family models cover 90+, others vary. Pick "auto-detect" or specify the language for highest accuracy.
Igama-impazamo ixabiso yi 5-10% kwi isiNgesi esicocekileyo sesandi, 10-20% kwisandi esinomsindo okanye esinomsindo. Iinketho ezinkulu zesi siqalelo sesandi zisebenza kakuhle kwiimeko ezinzima - khetha banzi xa isandi sinzima.
Ewe - icandelo ngalinye liquka isiqendu sokuqala/sokuphela kwexesha. Rhweba ngaphandle njenge-SRT okanye i-VTT kwaye ixesha lemaphu lihamba ngqo kwividiyo yakho.
Moonshine Base iqhuba kwiGPUs yethu ekhethekileyo echaphazela i-pool yakho ekhululekileyo yosuku kuqala; $5 → 200,000 i-token ehlawulweyo emva koko. malunga ne-~500 i-token ngomzuzu.
MP3, WAV, M4A, FLAC, OGG, kunye nevidiyo (MP4, MOV, WebM) - sikhupha isandi. Ubukhulu be-500 MB nganye. Iifayili ezide? Yahlula nge /audio/cut/ okanye sebenzisa /v1/stt/batch/.
Umthumeli we diarization udlula ngokuhlukileyo - tshintsha "udidi" kwi / transcribe /. Moonshine Base iphatha udidi; udidi luphawula icandelo ngalinye nomthumeli 1 / umthumeli 2 / njl.
Ewe - /iqela/ lixhasa isiqulathi seefayili zesandi. Ushicilelo ngalunye luya /i-akhawunti/?tab=imbali ngegama lefayili eliphambili. Ukugcina umthi wesiqulathi seefayili sebenzisa i-API.
Ewe - UTHENGA umsindo wakho kwi /v1/stt/transcribe/ ngemodeli "Moonshine Base". Ibuyisela i JSON ngombhalo + iinxalenye + i-word-level timestamp. /api/ ine-reference epheleleyo.
Iimodeli ezimkelweyo zigcina umsindo kwi-GPU yethu; i-premium idlula nge-DPA. Umsindo ususwa emva kwefestile yokwabelana (24h anon, 7d ungene). Asiyi kuqeqesha kwingxelo yakho.
Ewe - Free.ai inika ukusetyenziswa korhwebo kweencwadi ezibhaliweyo. Ufuna ilungelo lesandi olayishe (u Recording yakho, ilayisensi yezinto, okanye imixholo ngegunya).
Ixesha elipheleleyo liyi 0.05-0.2× - ipodcast yemizuzu engama-60 iguqula kancinane kwimizuzu engama-3-12. Iimodeli eziphezulu zihlala zigqiba ngokukhawuleza. Sebenzisa iqhosha lofolo ukuvala i-tab.