Ntuziaka Audio AI

Mimi gụgharia Audio Codec

Mimi bụ codec ọdịyo akwara nke na-etinye okwu n'ime obere iyi nke akara ngosi dị adị n'ezie, yabụ ụdị AI nwere ike ige ntị ma kwuo okwu na obere oge.

2 nkeji na-agụEmelitere ikpeazụ

Nchịkọta

It is the audio backbone behind Kyutai's Moshi voice model.

Ime miri emi

Mimi, nke ụlọ nyocha French Kyutai wepụtara na 2024, bụ codec neural nke na-atụgharị ụda 24 kHz ka ọ bụrụ iyi nke akara ngosi nwere ihe dị ka 1.1 kbps na naanị 12.5 tokens kwa nkeji. Ọ na-eji encoder-decoder nwere residual vector quantization (RVQ), na-ekewa akara n'ime 'semantic' ọkwa nke mbụ gbapụrụ site na ụdị okwu okwu na-elekọta onwe ya (WavLM) gbakwunyere ọtụtụ ọkwa 'acoustic' na-ejide udidi olu. N'ụzọ dị mkpa ọ na-enuba nke ọma na ihe kpatara ya: ọ na-ewepụta akara ka ọdịyo na-abịarute kama ichere vidiyo zuru ezu, yana ihe dị ka ms 80 nke latency. Nke a na-eme ka ụdị asụsụ na-emeso okwu dị ka akara ederede, na-enyere Moshi aka ịkparịta ụka na duplex zuru oke ma na-edobe ọdịyo arụgharịrị arụgharịrị nghọta na eke.

Nghọta nka nka

Aghụghọ Mimi bụ atụmatụ nkewa-RVQ. A na-azụ akwụkwọ koodu nke mbụ na nhụsianya nhụsianya ka ọ dakọtara na ntinye sitere na WavLM, na-amanye ya iburu 'pụtara' phonetic, ebe codebooks ụda olu na-emegharị nkọwa nke waveform. Ihe ntụgharị na-arụ ọrụ n'ime ọkpọ ahụ, na mfu nke mgbagha (GAN) na decoder na-amụba ogo mmepụta. Mgbagwoju anya ihe kpatara na-eme ka ihe niile na-aga n'ihu, yabụ latency na-anọ nso 80 ms.

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 Mimi Streaming Audio Codec

Na-atụ anya codecs dị ka Mimi ka ọ bụrụ ọkọlọtọ ọkọlọtọ n'etiti ọdịyo na ụdị asụsụ buru ibu, na-akpali ndị na-enyere aka olu oge n'oge nzaghachi sub-100 ms. Nchọpụta na-eme ka ọnụ ọgụgụ token dị ala ka ọ na-echekwa njirimara ọkà okwu, mmetụta, na egwu. N'ihi na Kyutai mepere Mimi na Moshi, ọ ga-abụ na ọ ga-amịpụta ọtụtụ sistemu okwu na-emepe emepe, ndị na-enyere aka na ngwaọrụ, yana ngwa nzikọrịta ozi olu bandwidth dị ala.

Mmejuputa n'ezie n'ụwa

Na-akwado onye enyemaka olu Duplex nke Kyutai's Moshi ka o nwee ike ige ntị na ikwu okwu n'otu oge

Ngosipụta akara okwu n'ụdị asụsụ maka ntụgharị okwu-gaa-okwu ozugbo

Oku olu dị obere bitrate (~ 1.1 kbps) maka ọnọdụ netwọkụ adịghị mma ma ọ bụ mkpọkọ

Ịkwado ụda olu maka imepụta okwu na pipeline ederede gaa n'okwu na-atụgharị uche dị ka ederede

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 Mimi Streaming Audio Codec 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

Codec ọdịyo akwara

Ajụjụ a na-ajụkarị

What is Mimi Streaming Audio Codec?

Mimi bụ codec ọdịyo akwara nke na-etinye okwu n'ime obere iyi nke akara ngosi dị adị n'ezie, yabụ ụdị AI nwere ike ige ntị ma kwuo okwu na obere oge. Ọ bụ ọkpụkpụ azụ ọdịyo dị n'azụ ụdị olu Moshi nke Kyutai.

Kedu ụlọ nyocha wepụtara Mimi na ụdị olu Moshi?

Ụlọ nyocha France Kyutai tọhapụrụ Mimi na Moshi na 2024, bụ nke meghere ha abụọ.

Ihe dị ka token ole ka Mimi na-ewepụta maka ikwu okwu?

Mimi na-echikota ọdịyo 24 kHz ka ọ bụrụ naanị ihe dị ka akara 12.5 kwa nkeji na ihe dị ka 1.1 kbps.

Kedu ihe koodu mbụ ('semantic') na Mimi weghara?

A zụrụ ọkwa RVQ nke mbụ site na distillation sitere na WavLM iji buru ọdịnaya semantic/phonetic, ebe ọkwa ọzọ na-agbakwunye nkọwa ụda.

Kedu ihe kpatara eji kọwaa Mimi dị ka 'gụgharia' ma ọ bụ 'ihe kpatara'?

Mimi na-ahazi ọdịyo n'ụzọ na-akpata ma na-ewepụta akara ngosi na-abawanye, na-enye ihe dị ka latency 80 ms dabara maka mkparịta ụka dị ndụ.

Kedu usoro ngụkọ ọnụ ka Mimi na-eji ezobe ọdịyo?

Mimi na-eji ọnụọgụ vector residual, na-edobe ọtụtụ codebook ka nke ọ bụla na-agbakwunye nkọwa kacha mma na nke gara aga.