Imisusire

i Gufungura - Inkomoko. Bihari Biturutse na Urubuga Imigaragarire.

Umwandiko- Kuri-

Chatterbox
Byemewe

Mu buryo bupfushije, twashoboye gukora ibishushanyo by'amajwi mu buryo bwa "zero-shot" kuva ku bushushanyo buvugana bugufi, mu rurimi rushika kuri 23. Modeli yacu y'ibanze yo gukora ibishushanyo by'amajwi, amajwi ashyirwa mu filimi n'ubusanzwe, n'imodeli yacu y'ibanze yo guhindura amajwi (Chatterbox VC).

Ururimi: 23
Igaragaza ry'amajwi Kuva ku minyago 6 kugeza ku minyago 60.
VRAM ~5GB
Inyandiko y'Iyemererakoresha MIT
Kugenzura ibintu bidatanga amakuru y'mbere. Ihinduramiterere Ururimi: Ikoresha:
, na

Imashini enye zishyira ijambo mu magambo (text-to-speech) zishyirwa hamwe mu mwirondoro umwe. Kokoro 82M (licensi ya Apache 2.0) ni yo yifashishwa cyane; Piper itanga imizinda gusa itanga ububasha bwo kuyikoresha mu buhamini; VITS yongera abavuga icyongereza 109, kandi Bark yongera ibice 130 mu rurimi 13.

Amashusho 316
Ururimi: 19
Amamotori 4
Iyimura MP3, WAV, OGG, FLAC. (Izi ni format z'amafilasi y'amajwi.)
Ururimi: Gukoresha
Umwandiko Kuri
OpenVoice v2

Umwanda w'ijwi n'imodeli yo gukoresha ijwi ry'undi. Ibyo bishobora gukorwa kandi, kandi muri iyi minsi, "Chatterbox VC" ni yo ikoreshwa cyane kuko itanga ijwi ry'umuntu ushaka gukoresha neza cyane.

Guhuza in Igerageza OpenVoice v2: 0.17 to 0.35
Guhuza in Igerageza Chatterbox VC: 0.64 to 0.75
Inyandiko y'Iyemererakoresha MIT

Ubwumvikane bw'amajwi amaze gutunganywa n'amajwi ahanditse y'umunyamigani. Ibyiciro by'amajwi by'umunyamigani umwe n'umwe byagaragaje ubumvikane bwa 0.76 kugeza 0.88.

Ihinduramiterere Bitari ngombwa
GitHub

Kuri Umwandiko

Whisper Large v3
Byakoreshejwe

Imodeli y'OpenAI yo gukoresha ibitekerezo by'abantu, ifite code ishyirwa ahagaragara, itumiza ijwi ry'amajwi. Itumiza amajwi mu rurimi rushika kuri 100, igashyira ibitekerezo by'igihe, kandi itanga ubweme bw'ururimi ku bushake.

Ururimi: 100
Ibigenga 1.5B
VRAM ~10GB
Ururimi: Ikoresha:
Urugero Mburabuzi Ururimi: Igihe- ngombwa
Whisper Large v3 Turbo

A, Verisiyo Bya V3 ya:. Nka -.

Ururimi: 100
Ibigenga 809M
VRAM ~6GB
Inyandiko y'Iyemererakoresha MIT
Neza. Ururimi:

Ikiganiro

Qwen2.5 7B Instruct

Gufungura - Inkomoko Ururimi: Urugero: i in & Ubutumwa, in i. Mucukumbuzi & Ubutumwa na A & Ubutumwa i & Ubutumwa.

Ibigenga 7B
Inyandiko y'Iyemererakoresha Apache 2.0
Igice Piper
ku Amakaradi yacu y'ishusho (GPUs).
Kugena
GitHub