Mgbasa ụdị StyleTTS 2
StyleTTS 2 bụ ụdị ederede gaa na-okwu nke na-emeso 'ụkpụrụ' olu - prosody, mmetụta, na timbre ọkà okwu - dị ka mgbanwe na-enweghị atụ nke ejiri ụdị mgbasa ozi, wee jikọta ọdịyo na ọzụzụ mmegide megide nnukwu ụdị asụsụ okwu.
Nchịkọta
It matters because it reached human-level naturalness on single-speaker benchmarks without needing a reference clip at inference time.
Ime miri emi
StyleTTS 2, nke ndị nyocha na Mahadum Columbia wepụtara na 2023, na-ewepụta okwu site n'ịtụle ụdị 'vector' nke latent site na iji usoro mgbasa ozi etinyere naanị na ederede ntinye, wee dezie ụdị ahụ gbakwunyere fọnịm n'ime usoro ifegharị. Ụdị vector na-achịkwa ihe niile edeghị n'ime ederede: ọnụ ọgụgụ okwu, contour innation, nkwụsịtụ, na agba mmetụta uche. N'ụzọ dị mkpa, ọ na-agbakwụnye ọzụzụ mmegide na nnukwu ụdị asụsụ a zụrụ azụ (WavLM) dị ka ndị ịkpa oke, na-akpali mmepụta ahụ gaa n'ezie ụda mmadụ na-ada ụda. Na benchmark LJSpeech ọ karịrị ndekọ mmadụ na ọkwa ndị na-ege ntị, yana na ọtụtụ ndị na-ekwu okwu LibriTTS setịpụrụ ya dakọtara eziokwu n'ala - ihe dị ịrịba ama maka àgwà TTS akwara njedebe na njedebe.
Nghọta nka nka
The isi aghụghọ bụ ụdị mgbasa: kama ịkọ otu ofu prosody, StyleTTS 2 ụdị style dị ka ihe gbasara nke puru omume nkesa na samples si na ya site a diffusion nlereanya na-agba ọsọ na a ala-akụkụ latent ohere, otú ahụ otu ahịrịokwu nwere ike na-ekwu ọtụtụ eke ụzọ. Ọgwụgwụ ruo n'ọgwụgwụ, amụma ogologo oge, ihe ngbanwe ụdị, decoder, na onye mmegide mmegide dabere na WavLM ka a zụrụ azụ, na-eme ka gradients na-esi na ogo waveform laghachi azụ site na pipeline dum.
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 StyleTTS 2 Mgbasa ụdị
Na-atụ anya mgbasawanye ụdị ka ọ ga-ejikọta ya na mmechi olu efu ka sekọnd ole na ole nke ntụnyere ọdịyo na-eduzi ụdị atụpụtara, yana iji njikwa njikwa nke na-ahapụ ndị okike ka ọ kpọọ mmetụta, mesie ike, ma ọ bụ ọsọ ọsọ n'ụzọ doro anya. Ụdị dị nro dị ọkụ na-achọ igbutu nlele mgbasa ozi ọtụtụ nzọụkwụ maka iji ozugbo na ngwaọrụ. Ka ụdịdị ndị a na-erute ogo mgbasa ozi, akara mmiri na nkwenye nkwenye ga-abụ ọkọlọtọ iji lebara nchegbu nkwubi okwu na-ezighị ezi anya.
Mmejuputa n'ezie n'ụwa
Ịmepụta akwụkwọ akụkọ ọdịyo ebe otu ọkà okwu na-adịgasị iche iche n'ofe isiakwụkwọ kama ị na-ada ụda monotone
Na-ewepụta olu agwa agwa maka egwuregwu indie na animation na-eweghị ọtụtụ ndị na-eme olu olu
Ndị na-agụ ihuenyo ohere ịnweta ike nke na-ada ụda mmadụ zuru oke maka ige ntị ogologo oge
Ịmepụta ụda olu mmụta e-mmụta mpaghara nwere nkwusi ike sitere n'okike na imegharị site na ederede ederede dị larịị
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
Nweta nkwenye doro anya maka ijide olu, imechi, na ijigharị.
Nwale ogo n'ofe ndị na-ekwu okwu dị iche iche yana ọnọdụ ndabere.
Kọwaa mgbe mmadụ ga-enyocha ma ọ bụ kwado nsonye.
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 StyleTTS 2 Style Diffusion quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Ntuziaka na-esote
Mgbasa spectrogram Riffusion
Ajụjụ a na-ajụkarị
What is StyleTTS 2 Style Diffusion?
StyleTTS 2 bụ ụdị ederede gaa na-okwu nke na-emeso 'ụkpụrụ' olu - prosody, mmetụta, na timbre ọkà okwu - dị ka mgbanwe na-enweghị atụ nke ejiri ụdị mgbasa ozi, wee jikọta ọdịyo na ọzụzụ mmegide megide nnukwu ụdị asụsụ okwu. Ọ dị mkpa n'ihi na ọ ruru ogo mmadụ na akara nrịbama otu okwu na-achọghị obere nrụtụ aka n'oge ntụgharị.
Kedu ihe ụdị StyleTTS 2 na-eji usoro mgbasa ozi?
StyleTTS 2 na-atụpụta vector ụdị dị ala nwere ụdị mgbasa ozi, na-ewere njirimara prosody, mmetụta, na njirimara ọkà okwu nke ederede akọwapụtaghị.
Kedu ụdị okwu a zụrụ azụ nke a na-eji dị ka onye mmegide mmegide na StyleTTS 2?
StyleTTS 2 na-eji ụdị asụsụ okwu buru ibu dị ka WavLM dị ka ndị ịkpa oke na ọzụzụ mmegide iji kwalite nsonaazụ na ọdịyo mmadụ.
Kedu ihe kpatara StyleTTS 2 nwere ike isi kwuo otu ahịrịokwu n'ọtụtụ ụzọ eke?
N'ihi na a na-ewere ụdị dị ka ihe na-agbanwe agbanwe ma na-eṅomi site na mgbasa ozi, ọgbọ ọ bụla nwere ike ịmepụta ihe dị iche iche ma ọ bụ eke maka ederede.
Kedu akara nrịbama otu onye na-ekwu okwu ka StyleTTS 2 kọrọ karịa ndekọ mmadụ na ọkwa ndị na-ege ntị?
Na akara nrịbama nke LJSpeech, StyleTTS 2 nwetara ọkwa ịdị mma nke ndị na-ege ntị karịa ndekọ nke ala-eziokwu mmadụ.
Kedu ihe ọzụzụ 'njedebe-na-ọgwụgwụ' na-enye StyleTTS 2?
Ọzụzụ ngwụcha ngwụcha ngwụcha na-eme ka mfu na ogo ọdịyo ikpeazụ kwalite amụma ogologo oge, ihe ntinye ụdị, na decoder ọnụ karịa n'ụdị dịpụrụ adịpụ.