Ingxelo ye Bug / Isicelo se Feature

Ixesha elipheleleyo le TTS

Uthutho lombhalo-ukuthetha-ukuthetha ngexesha elifutshane lesibini-lesibini lesandi. Ifakwe kubameli belizwi kunye neenkqubo eziphilayo.

Asikho nasiphi na isandi se-TTS kwisiNgesi sakho. Nceda uncedo lwethu ukongeza isandi sakho! Intengiso yelizwi lakho

& Umbhalo

Unikezelo
0/5,000 Iimpawu ~0.3s I-audio yokuqala

Izicwangciso Zesandi

Iimodeli ezikwaziyo ukudlulisa kuphela.

Ixesha elimiselweyo eliphilayo

Cofa i-Stream ukulinganisa i-latency yesandi yokuqala

Imveliso

Ii-chunks zesandi ziya kudlala apha xa ziphuma.

0:00
Ingxenye yokuqala:
Iinxalenye ezipheleleyo: 0
Ixesha elipheleleyo:

Indlela i-TTS esebenza ngayo

1. Thumela Umbhalo

Umbhalo we-POST kwi /v1/tts/stream/ njengesicelo seSeva-Ethunyelwe Iziganeko.

2. Imodeli ivelisa

I-Kokoro iquka umbhalo kwaye ivelisa iiseti zesandi ngesethi kwi-GPU.

3. IiNkqubo Zomthotho

I-Base64-encoded WAV chunks ifika ngaphezulu kwe-SSE kwaye iqala ukudlala ngokuzenzekelayo.

4. Listen Live

Umsebenzisi ufumana isiqalo sombhalo ngaphantsi kwesekondi, nokuba kungeniso olude.

Iimeko Zokusetyenziswa

apho i-sub-second latency ivula iimvakalelo ezintsha.

IiNkqubo zeSandi

Ii-bots zonxibelelwano eziphendula ngokukhawuleza njengoko umntu angaphendula.

Ukuphinda usebenze ngokuzenzekelayo

Gcina ifayile ye JSON

Imidlalo

Incoko yababini ye-NPC ephendula ngokuzenzekelayo kukhetho lomdlali, akukho VO eyenziwe ngaphambili.

Ufikelelo

Abafundi bekhusi kunye neezixhobo ezincedisayo eziqala ukuthetha xa umsebenzisi ecofa.

Iinkqubo ze TTS zexesha elibonakalayo

Qala ngokukhululekileyo, uphucule xa ufuna okuninzi

Iinketho zelizwe
  • Kokoro streaming (imodeli ekhululekileyo)
  • 500 iimpawu nganye
  • 10 ii-streams ezikhululekileyo/imini nganye kumsebenzisi ongaziwayo
  • Ixesha elimiselweyo lesandi lokuqala lesibini
  • SSE isasaza ngaphezulu kwe-HTTPS
Ethandwa Kakhulu
I-akhawunti Ekhululekileyo
  • 15, 000 iimpawu xa ubhalisa
  • 5,000 iimpawu ngomjelo ngamnye
  • Isitshixo se-API sokufikelela ngenkqubo
  • Imbali yobuhlanga
  • Akukho xesha lokuphuma ligqityiwe
Ubhaliso
I-Pro
  • MOSS-TTS-Realtime (xa uphila)
  • 100, 000 iimpawu ngomjelo ngamnye
  • Ufolo lwe-GPU oluphambili
  • I-voice agent + Twilio integration
  • Iimali ezihlawulwe
Yenza phezulu

Imibuzo ebuzwa rhoqo

Ixesha- lexesha elipheleleyo lombhalo- ukuya- ku- kuthetha udlulisa iichunks zesandi njengoko zidalwa, endaweni yokulinda ilizwi lonke ligqibeke. Isampulu yokuqala yesandi ifika ngaphantsi kwesekondi enye, iyenza ilungele abameli besandi abaphilayo, ukudubula, kunye neenkqubo ezisebenza ngokudibeneyo apho ukungabikho kwexesha libalulekile.

I-TTS eqhelekileyo ivelisa ifayile yesandi epheleleyo phambi kokuba ibuyisele nantoni na - ulinde, ngoko uthethe umgca wonke ngaxeshanye. Ixesha lexesha elipheleleyo i-TTS isebenzisa Iziganeko ezithunyelwe ngumncedisi (SSE) ukudlulisa iichunks zesandi ezimfutshane njengoko imodeli ivelisa. Umsebenzisi uthetha ukuqala komgca ngokukhawuleza, nakwingeniso elide.

I-Kokoro yindawo emiselweyo yokumva — ivelisa isandi esimalunga ne-100x esikhawulezayo kunexesha elibonakalayo kwi-GPU yakudala. Sidibanisa i-MOSS-TTS-Realtime njengento engcono kakhulu; abasebenzisi baya kuba nakho ukukhetha ngesicelo ngasinye xa ifika.

Ixesha eliqhelekileyo lesandi sokuqala kwiKokoro liyi-300-800ms ngaphezulu konxibelelwano lwabucala. Uthungelwano lwe-round-trip lulawula emva koko. Iphepha libonisa ixesha elipheleleyo elilinganiselwe kwisandi sokuqala kwi-UI ukuze ubone ngokuchanekileyo ixesha elide elithatha isicelo ngasinye.

Ii-agents zesandi eziphendula ngokuncokola, ukuphinda ubhale ngokuzenzekelayo umxholo osasazwayo, i-NPCs yomdlalo osebenza ngokuzenzekelayo, abafundi bokufikelela abaqala ukuthetha xa umsebenzisi ecofa, nakweyiphi na isicelo apho ulinde iiyure ezimbini okanye ezintathu zesandi kuziva kunzima.

Ewe. UTHENGA ku https://api.tts.ai/v1/tts/stream/ ngequbuliso elifanayo njenge /v1/tts/ isiqendu esiqhelekileyo. Impendulo yi SSE umbhobho we base64-encoded WAV chunks. Umphakamo okhululekileyo uxhasa iindidi ezili-10 ngosuku ngalunye kumsebenzisi ongaziwayo; abasebenzisi abaqinisekisiweyo bafumana unikezelo olupheleleyo lwe-akhawunti nganye.

I-Kokoro isebenzisa ilizwi eliqeqeshiweyo kwaye alifani. I-MOSS-TTS-Realtime (xa idityaniswe) ixhasa ukufana kwelizwi elijikelezayo ukusuka kwi-3-second reference. Ukufana kwelizwi elipheleleyo namhlanje, sebenzisa iphepha eliqhelekileyo /text-to-speech/ nge-Chatterbox okanye i-GPT-SoVITS - ezi azikwazi ukudlulisa kodwa zivelisa ilizwi elikhethekileyo.

Ixabiso le-character elifanayo njenge-TTS eqhelekileyo yendawo yesiphelo. I-Kokoro i-free-tier (1x ixabiso). I-MOSS-TTS-Realtime izakusebenza kwi-standard-tier (2x ixabiso) xa yenziwe. Inkqubo yokusebenza yokusasaza ayifakwanga nayiphi na ixabiso elingaphezulu.

Ewe — dibanisa indawo yokugqitha yokusasaza ne-Twilio voice webhook ukunika umculo ophilayo kwintlanganiso yefowuni. Inkqubo yethu ye-voice agent sele yenza oku kwi-IVR kunye noqhagamshelwano oluphumayo. Ukuphuma kwexesha lokulibaziseka kwefowuni kufutshane ne-end-to-end kuqhelekileyo kukwimizuzu emi-1-2 kubandakanya i-STT kunye ne-LLM response.

Ukuba uthungelwano lwakho lushiya inxalenye ehambayo, umdlali osasazayo uya kuhamba phambili kunokuba uhlale. Kwizicelo ezingazithobeli izithuba, buyela umva kwinqanaba lokugqitywa eliqhelekileyo elingenaso isiqhagamshelanisi, okanye ugcine i-500ms yesandi phambi kokuba uqale ukudlala.
5.0/5 (1)

Yintoni esinokuyilungisa? Ulwazi lwakho olufunyenweyo lunceda silungise iingxaki.

Ulwazi lwe KDEName

Isimahla kwiminyaka emi-10 yokuqala yosuku. Bhalisa ukuvula unikezelo olupheleleyo lophawu kunye nokufikelela kwi-API.