Models
Anyị na-ejikọta ihe kacha mma na-emeghe-isi okwu AI models. All models are available via API and web interface.
Nhazi ngwe-ka-asụsụ
Chatterbox
E akwadoZero-shot ụda cloning site na a obere reference sample, na 23 asụsụ. Anyị difọ́ọ̀ltụ̀ cloning model, ụda na AI dubbling, na anyị difọ́ọ̀ltụ̀ ụda ntụgharị model (Chatterbox VC).
Kokoro, Piper, VITS na Bark
Nwa na-emepe ngwe na-emegharị n'ime okwu engines n'azụ otu n'ime ndị na-ahọrọ olu. Kokoro 82M (Apache 2.0) bụ difọ́ọ̀ltụ̀; Piper na-enye naanị ụda nke data nlekọta na-enye ohere iji ahịa; VITS na-agbakwunye 109 English speakers na Bark 130 presets na asụsụ 13.
OpenVoice v2
Tone agba-ntụgharị na kloo model. Oge ọ bụla a ga-ahọrọ maka ụda ntụgharị, ebe Chatterbox VC bụ ugbua difọ́ọ̀ltụ̀ n'ihi na ọ na-echekwa ọtụtụ ụda n'ime n'ime.
Speaker similarity of converted recordings to a different clip of the target speaker. Two clips of the same real speaker score 0.76 to 0.88.
Nsụgharị
Whisper Large v3
Ejirila ya ọtụtụOpenAI's open-source speech recognition model. Na-atụgharị asụsụ 100, na oge-eziga na nchọpụta asụsụ nkeonwe.
Whisper Large v3 Turbo
A obere, ngwa ngwa nsụgharị nke Whisper Large v3 maka ngwa ngwa transcripts. Họrọ ya dịka whisper-turbo.
Nchedo ụda
Qwen2.5 7B Instruct
Open-source asụsụ móòdù nke na-edepụta nzaghachi na nchọgharị ahụ, n'ime onye ị họọrọ. Nchọgharị gị na-edepụta okwu gị na ụda Piper na-edepụta nzaghachi n'ụzọ dị elu.