Audio AI Itọsọna

AudioLM

AudioLM jẹ ilana iwadii Google ti o ṣe agbejade ohun ojulowo — ọrọ tabi orin piano — nipa ṣiṣe itọju ohun bii ede ati asọtẹlẹ ami-ami nipasẹ ami-ami.

2 min kakẹhin imudojuiwọn

Akopọ

It matters because it showed you can produce coherent, natural-sounding audio continuations without any text transcript or musical score.

Jin Dive

Iṣagbekale nipasẹ Google ni ọdun 2022, AudioLM ṣe atunṣe iran ohun afetigbọ bi iṣoro awoṣe-ede: o ṣe iyipada awọn ọna igbi aise sinu awọn ami iyasọtọ ati lẹhinna sọ asọtẹlẹ ami atẹle, gẹgẹ bi awoṣe ọrọ ṣe asọtẹlẹ ọrọ ti nbọ. Ẹtan bọtini rẹ jẹ ilana-iṣe ti awọn oriṣi ami. Awọn ami ami 'Semantic' (lati awoṣe bii w2v-BERT) gba eto igba pipẹ - phonetics, syntax, orin aladun - lakoko ti awọn ami 'acoustic' (lati SoundStream neural codec) gba awọn alaye to dara bi idanimọ agbọrọsọ, timbre, ati awọn ipo gbigbasilẹ. Nipa sisọ asọtẹlẹ awọn ami isọdi akọkọ, lẹhinna mimu awọn ami ami akositiki sori wọn, AudioLM ṣe agbejade awọn itesiwaju ti o duro ni isọdọkan fun awọn aaya pupọ lakoko titọju ohun atilẹba tabi ohun elo. Fun iṣẹju diẹ ti ọrọ, o tẹsiwaju ni sisọ ni ohun kanna; fun piano, o improvises ni kanna ara.

Imọ-imọ-ẹrọ

AudioLM ti ni ikẹkọ nikan lori ohun - ko si awọn iwe afọwọkọ. SoundStream fun ohun afetigbọ sinu awọn ami akositiki nipasẹ pipọ pikito ti o ku, lakoko ti w2v-BERT n pese awọn ami atunmọ isokuso. Akopọ ti awọn awoṣe ede Ayirapada sọ asọtẹlẹ awọn ami ni awọn ipele: itumọ akọkọ fun igbekalẹ, lẹhinna isokuso ati awọn ami akositiki itanran fun atunkọ-iṣotitọ giga. Oluyipada SoundStream nikẹhin yi awọn ami isọtẹlẹ pada si ọna igbi kan, ohun afetigbọ ti o jẹ ki ohun agbọrọsọ jẹ ki o ṣe deede.

Ipa Ilana

Wiwọle ati arọwọto

O ṣe ilọsiwaju iraye si nipasẹ transcription, alaye, ati awọn atọkun ohun.

Iye owo ati isuna

Awọn ẹgbẹ Media le firanṣẹ ohun didan yiyara pẹlu awọn isuna-owo kekere.

Iyara ati iwọn

Awọn ọna ṣiṣe ti nkọju si alabara le ṣe ilana awọn ibaraẹnisọrọ sisọ ni iwọn nla.

Ojo iwaju ti AudioLM

Ohunelo ti o da lori ami-ami AudioLM di ipilẹ fun awọn ọna ṣiṣe atẹle: Google's Awọn imọran AudioLM ti a jẹ sinu MusicLM fun orin-si-orin ati SoundStorm fun iran yiyara, lakoko ti aaye ti o gbooro ni bayi dapọ awọn ami atunmọ ati awọn ami akustic kọja ọrọ, orin, ati awọn ipa ohun. Reti yiyara, iran-akoko gidi, awọn abajade ibaramu gigun, ati iṣakoso multimodal nibiti ọrọ tabi awọn ifihan agbara miiran ti darí awọn awoṣe ikẹkọ ohun-orin lasan. Awọn imọ-ẹrọ kanna tun mu awọn ifiyesi pọ si nipa pipọ ohun ati awọn jinlẹ ohun.

Real-World imuse

Tesiwaju agekuru ọrọ kukuru ni ohun agbọrọsọ kanna ati intonation laisi iwe afọwọkọ kan

Imudarasi orin duru tuntun ti o baamu ara ti itusilẹ kukuru ti o gbasilẹ

Ṣiṣẹ bi ẹhin-iran ohun ohun fun awọn ọna ṣiṣe orin-si-orin bii MusicLM

Iwadi sinu iṣelọpọ ọrọ ti o ṣe itọju prosody ati awọn acoustics gbigbasilẹ lati inu apẹẹrẹ kan

Awọn ewu & Awọn ọna iṣọ

ilokulo ohun ati awọn ewu afarawe ṣe pọ si nigbati igbanilaaye ba sonu.

Yiye le ju silẹ kọja awọn asẹnti, awọn ede-ede, tabi awọn agbegbe alariwo.

Ohun afetigbọ sintetiki le jẹ aṣiṣe fun ọrọ ododo laisi isamisi to yege.

Ilana Ilana imuse

1

Gba ifọkansi ti o fojuhan fun gbigba ohun, ti ẹda, ati ilotunlo.

2

Didara idanwo kọja awọn agbohunsoke oniruuru ati awọn ipo abẹlẹ.

3

Ṣetumo nigbati eniyan gbọdọ ṣe atunyẹwo tabi fọwọsi awọn abajade.

4

Aami ohun sintetiki ki o tọju awọn igbasilẹ provenance fun iṣiro.

Tesiwaju Ṣiṣawari

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 AudioLM quiz

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

Bẹrẹ adanwo

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

Itọsọna atẹle

Koko Spotting ati Wake Words

Awọn ibeere ti a beere nigbagbogbo

What is AudioLM?

AudioLM jẹ ilana iwadii Google ti o ṣe agbejade ohun ojulowo — ọrọ tabi orin piano — nipa ṣiṣe itọju ohun bii ede ati asọtẹlẹ ami-ami nipasẹ ami-ami. O ṣe pataki nitori pe o fihan pe o le gbejade isomọ, awọn itesiwaju ohun afetigbọ ohun adayeba laisi eyikeyi iwe afọwọkọ ọrọ tabi Dimegilio orin.

Ero pataki wo ni AudioLM lo lati ṣe agbejade ohun?

AudioLM ṣe iyipada awọn fọọmu igbi sinu awọn ami iyasọtọ ati sọtẹlẹ wọn lẹsẹsẹ, ni lilo ọna ami atẹle ti awọn awoṣe ede si ohun aise.

Kini ipa ti awọn ami 'semantic' ni AudioLM?

Awọn ami atunmọ (lati w2v-BERT) ṣe koodu igbekalẹ ipele giga ati isọdọkan, lakoko ti awọn ami akositiki mu awọn alaye ohun afetigbọ to dara.

Kodẹki nkankikan wo ni o ṣe agbejade awọn ami akositiki AudioLM?

SoundStream nlo iye iwọn fekito ti o ku lati funmorawon ohun sinu awọn ami akositiki ati nigbamii decodes awọn ami asọtẹlẹ pada sinu fọọmu igbi kan.

Kini AudioLM nilo bi data ikẹkọ?

Ẹya ti o ṣe akiyesi ni pe AudioLM ṣe ikẹkọ lori ohun afetigbọ nikan laisi awọn aami ọrọ eyikeyi, eto ẹkọ taara lati ohun.

Kini idi ti AudioLM ṣe asọtẹlẹ awọn ami atunmọ ṣaaju awọn ami akositiki?

Ti o npese isokuso be akọkọ ntọju awọn o wu ni ibamu lori akoko; awọn ami akositiki lẹhinna ni iloniniye lori eto yẹn lati ṣafikun alaye iṣootọ giga.