Ntuziaka Audio AI

Okwu Moshi Full-Duplex

Moshi bụ isi okwu mepere emepe, ezigbo oge AI sitere na Kyutai na-ekwu okwu ma na-ege ntị n'otu oge - duplex zuru oke - kama ịtụgharị uche.

2 nkeji na-agụEmelitere ikpeazụ

Nchịkọta

That removes the awkward lag and rigid turn-taking of traditional voice assistants.

Ime miri emi

Moshi, nke ụlọ nyocha French Kyutai wepụtara na 2024, bụ ihe atụ ntọala okwu na-okwu nke e wuru maka mkparịta ụka ebumpụta ụwa, nke dị ala. N'adịghị ka ndị enyemaka pipeline na-eji okwu-gaa-ederede, mgbe ahụ ụdị asụsụ, wee ederede-gaa-na-okwu, Moshi na-ejikwa ọdịyo ozugbo na-aga n'ihu. Isi echiche ya bụ duplex zuru oke: ọ na-emepụta iyi iyi egwu abụọ n'otu oge - nke onye ọrụ na nke ya - ya mere ọ nwere ike ige ntị mgbe ọ na-ekwu okwu, jikwaa nkwụsịtụ, ọwa azụ nwere 'mhm' ma na-agafe n'ụzọ nkịtị dị ka ụmụ mmadụ na-eme. Ọ na-erute latency gburugburu 160-200 milliseconds, dị n'okpuru ọnọdụ enyemaka enyemaka. N'okpuru mkpuchi ọ na-ejikọta ụdị ederede na ọdịyo 7B-parameter (Helium) na Mimi, codec ọdịyo akwara nke na-etinye okwu n'ime akara ngosi nwere ike ịmepụta. Kyutai weputara nha na koodu n'ihu ọha.

Nghọta nka nka

Aghụghọ Moshi bụ codec Mimi ya, nke na-atụgharị ọdịyo na-aga n'ihu ka ọ bụrụ iyi nwere obere ntakịrị nke akara ngosi dị na 12.5 Hz, gụnyere akara semantic na-agbanwe agbanwe. Ụdị asụsụ ahụ na-ebu amụma ihe nrịbama okwu nke ya na nke onye ọrụ n'otu oge iyi, yabụ ọgbọ agaghị akwụsị 'ige ntị'. Usoro 'Inner Monologue' na-ebu amụma ederede tupu odiyo, na-emeziwanye ogo asụsụ yana nkwekọ nke ihe Moshi na-ekwu n'ezie.

Mmetụta atụmatụ

Nweta na iru

Ọ na-eme ka nnweta ya dịkwuo mma site na ndegharị, ịkọ akụkọ, na ntụgharị olu.

Ọnụ ego na mmefu ego

Ndị otu mgbasa ozi nwere ike ibubata ọdịyo a na-egbu maramara ngwa ngwa site na iji obere mmefu ego.

Ọsọ na ọnụ ọgụgụ

Sistemụ na-eche ihu ndị ahịa nwere ike hazie mkparịta ụka n'ọtụtụ buru ibu.

Ọdịnihu nke Moshi Full-Duplex Okwu

Nlereanya duplex zuru oke na-aghọ ndebiri maka ụda olu AI, na-emetụta sistemụ n'ofe ụlọ ọrụ ahụ. Na-atụ anya obere, ụdị na ngwaọrụ, nkwado ọtụtụ asụsụ, obere oge, na ntinye n'ime ndị nnọchite, ọrụ ndị ahịa, na ngwa nnweta. N'ihi na Moshi mepere emepe, ndị nchọpụta nwere ike nyochaa ma melite ya n'efu. Ihe ịma aka na-adịgide gburugburu ntụkwasị obi n'eziokwu, nchekwa n'okwu ọnụ, yana nhụsianya mmetụta uche, mana ngbanwe site na ngbanwe siri ike gaa na mkparịta ụka na-adịghị akwụsị akwụsị nwere ike na-adịgide adịgide.

Mmejuputa n'ezie n'ụwa

Onye otu olu na-enweghị aka ị nwere ike ịkwụsị n'etiti ahịrịokwu, na nzaghachi na-erughị 200 milliseconds.

Mepee ntọala nyocha maka ịmụ mkparịta ụka n'ezie, duplex zuru oke na-enweghị igbe ojii.

Ndị enyemaka nnweta na-akparịta ụka n'ezoghị ọnụ na ndị ọrụ chọrọ ọsọ ọsọ, azụ na-apụta.

Na-egosipụta bots olu ọrụ ndị ahịa na-akwụsị akwụsị na-emeghachi omume mgbe onye na-akpọ oku ka na-ekwu okwu.

Ihe ize ndụ & okporo ụzọ nche

Iji olu eme ihe na ihe egwu mpụta ga-abawanye mgbe nkwenye na-efu.

Izi ezi nwere ike ịdaba n'ofe ụda olu, olumba ma ọ bụ gburugburu mkpọtụ.

Enwere ike imehie ọdịyo sịntetik dị ka ezigbo okwu na-enweghị akara doro anya.

Map mmejuputa

1

Nweta nkwenye doro anya maka ijide olu, imechi, na ijigharị.

2

Nwale ogo n'ofe ndị na-ekwu okwu dị iche iche yana ọnọdụ ndabere.

3

Kọwaa mgbe mmadụ ga-enyocha ma ọ bụ kwado nsonye.

4

Deba aha ọdịyo sịntetik ma debe ndekọ ihe ndekọ maka ịza ajụjụ.

Nọgide na-eme nchọpụta

Free newsletter

Get the daily AI briefing

Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.

One email each weekday. Unsubscribe in one click. We never sell or share your address.

Test yourself

Take the Moshi Full-Duplex Speech quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Malite ajụjụ

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Ntuziaka na-esote

Nhazi ederede maka okwu

Ajụjụ a na-ajụkarị

What is Moshi Full-Duplex Speech?

Moshi bụ isi okwu mepere emepe, ezigbo oge AI sitere na Kyutai na-ekwu okwu ma na-ege ntị n'otu oge - duplex zuru oke - kama ịtụgharị uche. Nke ahụ na-ewepụ ọdịghị ihe na-adịghị mma na nkwụsi ike nke ndị enyemaka olu ọdịnala.

Kedu ihe 'duplex' pụtara n'ihe gbasara Moshi?

Duplex zuru ezu pụtara na ihe nlereanya a na-ejikwa ụda na-abata na nke na-apụ apụ n'otu oge, ya mere ọ nwere ike ige ntị mgbe ọ na-ekwu okwu ma na-edozi nkwụsịtụ n'onwe ya.

Kedu otu wepụtara Moshi?

Kyutai, ụlọ nyocha nyocha nke French AI, mepụtara Moshi na 2024.

Kedu ihe bụ Mimi na usoro Moshi?

Mimi bụ codec ọdịyo akwara nke na-ejikọta okwu na-aga n'ihu n'ime iyi obere bitrate nke akara ngosi nwere ike iwepụta.

Kedu ka Moshi si dị iche na ọkpọkọ olu enyemaka olu ọdịnala?

Ndị enyemaka ọdịnala yinye okwu-gaa-ederede, ụdị asụsụ, na ederede-gaa-na-okwu; Moshi bụ otu ụdị okwu-gaa-na-okwu na-ejikwa ọdịyo na-aga n'ihu.

Kedu ihe mwepu mkparịta ụka ka Moshi lekwasịrị anya?

Moshi na-enweta latency n'ihe dị ka 160-200 ms, dị ala karịa ndị na-enyere aka olu na-ahụkarị, na-enye ohere ndakọrịta na nkwụsị nke okike.