faster-whisper large-v3
Free.ai (self-hosted)
·
stt
·
~500 Koffice FilterName minute
faster-whisper large-v3 is a spraak- to- text model gebou deur OpenAI / SYSTRAN. Sterkste by Accurate transcription. Self-hosted on Free.ai GPUs — runs free against your daily token pool (500 tokens per minuut). Released under MIT — commercial use permitted on Free.ai.
Gebruik via API
OpenAI- versoenbaar met REST API. Genereer 'n sleutel en noem hierdie model in sekondes.
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://..."}'Vrae wat dikwels gevra word
faster-whisper large-v3 trangeskryfte oudio in teks. Oplaai 'n MP3, WAV, M4A, of videolêer en faster-whisper large-v3 gee terug die volledige transkripsie plus opsionele SRT/VT onderregte met tyetampe.
faster-whisper large-v3 handles dozens of languages — Whisper-family models cover 90+, Parakeet covers ~25, others vary. Pick "auto-detect" or specify the language for highest accuracy.
Word-error rate is 5–10% on clean English audio, 10–20% on noisy or accented audio. Large variants of the same architecture do meaningfully better on hard cases — pick larger when the audio is rough.
Ja ☞ elke segment sluit in begin/end tyetampe. Voer uit as SRT of VTT en die kere kaart reguit op jou video.
faster-whisper large-v3 loop eers op ons eie GPU's teen jou daaglikse gratis swembad; R2 500 het per minuut daarvoor geskenk.
MP3, WAV, M4A, FLAC, OG, plus video (MP4, MOV, WebM) ons onttrek die oudio. Max 500 mb per oplaai. langer lêers? Skei met /audio/cut/ of gebruik /v1/st/bat/.
Luidspreker diarisering is 'n aparte herhaling ą skakel "diariseer" op /trantint /. faster-whisper large-v3 hanteer die transkripsie; diarisering etikette elke deel met Luidder 1 / Speaker 2 / etc.
Ja verblyd /bat/ aanvaar 'n gids van oudio lêers. Elke transkripsie lande in /accote/?tab=historie met die oorspronklike lêernaam. Vir kabinet- boom bewaar gebruik die API.
Ja KLAS POST jou oudio na /v1/strat/trant/ met model="faster-whisper large-v3". Gee terug JSON met teks + segmente + woord-vlak maaltampe. /api/ het die volledige verwysing.
Self-gehostde modelle hou klank op ons GPUs; premasie gaan deur met 'n DPA. Audio word uitgevee na die deel-venster (24h anon, 7d getekende-in). Ons oefen nie op jou invoers nie.
Yes — Free.ai grants commercial use of transcripts. You need rights to the audio you uploaded (your own recording, licensed material, or content with consent).
Real-time factor is roughly 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.