Moonshine Base
Free.ai (self-hosted)
·
stt
·
~500 tokens per minute
Moonshine Base ke a Mofuta oa puo ho ea ho mongolo e hahiloeng ke Useful Sensors. E ka ba e le Low-latency live transcription, embedded devices.. E na le li-Free.ai GPUs tse nang le li-server tse 100 tse nang le li-server tse 100 tse nang le li-server tse 100. E lokollotsoe ka MIT — ho sebelisoa ka khoebo ho lumelloa ka Free.ai.
Ho sebelisa ka API
REST API e lumellanang le OpenAI. E etsa konopo'me e bitsa mofuta ona ka metsotsoana.
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://..."}'Litšoantšo tse tšoanang
Lipotso tse botsoang khafetsa
Moonshine Base e ngola molumo o boleloang ka ho ba lengolo. Kopitsa faele ea MP3, WAV, M4A, kapa video'me Moonshine Base e tla u khutlisa lengolo le felletseng le lihlooho tsa SRT/VTT tse sa tšoaneng le li-timestamps.
Moonshine Base e sebetsana le lipuo tse ngata — Whisper-family models cover 90+, Parakeet covers ~25, others vary. Pick "auto-detect" or specify the language for highest accuracy.
Theko ea phoso ea mantsoe ke 5-10% ka audio ea Senyesemane e hloekileng, 10-20% ka audio e nang le mohala kapa e nang le mohala. Liphetoho tse kholo tsa ts'ebetso e ts'oanang li etsa hore li be betere ka liketsahalo tse thata - khetha tse kholo ha audio e le thata.
Ee — karolo e ngoe le e ngoe e na le li-timestamps tsa ho qala/ho fela. E romelloa ka SRT kapa VTT'me li-timestamps li tlatsoa kapele ho video ea hau.
Moonshine Base e sebetsa ka GPUs ea rona e le 'ngoe ho ea ka pool ea hau ea mahala ka letsatsi la pele; $ 5 → 200,000 tokens e tšehelitsoeng ka mor'a moo. Ka li-tokens tse ka bang 500 ka metsotsoana.
MP3, WAV, M4A, FLAC, OGG, le video (MP4, MOV, WebM) - re tlosa audio. Max 500 MB ka ho kenya. Lifaele tse telele? Tsebisoa ka /audio/cut/ kapa sebelisa /v1/stt/batch/.
Li-diaries tsa motsamaisi ke litsela tse fapaneng — toggle "diarize" ho /transcribe/. Moonshine Base e sebetsana le ho ngola; li-diaries li tšoaea li-segments ka ho toba le Motsamaisi 1 / Motsamaisi 2 / jj.
Ee — /batch/ e amohela lebokose la lifaele tsa audio. Lingoloa tsohle li tla ba /account/?tab=history le lebitso la faele la mantlha. Ho boloka moqhaka oa lebokose la lifaele sebelisa API.
E-na — POST audio ea hau ho /v1/stt/transcribe/ le model="Moonshine Base". E khutlisa JSON le mongolo + li-segments + timestamps tsa boemo ba mantsoe. /api/ e na le ho ngolisoa ka botlalo.
Li-model tse sebete li boloka molumo ka GPU ea rona; premium e tsamaea ka DPA. Molamu o tlosoa ka mor'a ho arolelana-fensetereng (24h anon, 7d e kentsoeng). Re sa koetlise ka li-input tsa hau.
Ee — Free.ai e fana ka ts'ebeliso ea khoebo ea li-transcripts. U hloka litokelo tsa audio eo u e kentseng (rekoto ea hau, thepa e nang le laesense, kapa litaba ka tumello).
Nako ea nako e ka bang 0.05-0.2× - podcast ea metsotso e 60 e ngola metsotso e 3-12. Li-models tsa Premium li ka qetella ka potlako. Sebelisa konopo ea lebokose ho koala tab.