I-Kokoro

Standard

I-lightning-fast, lightweight TTS with natural quality

Very Fast Isivinini
Good Ubunjani
Akukho Ukuklonya
9 Izilimi

Ngo I-Kokoro

802.11b/g/n, 802.11g/n, 802.11g/n, 802.11g/n, 802.11g/n, 802.11g/n, 802.11g/n

Izici ezibalulekile

Isisindo esincane kakhulu

82M parameters, ~300MB model size. Runs on CPU with minimal resources.

Isikhathi esincane

Ikhiqiza amagama asheshayo kunalezo ezidlalwa ngokushesha, ngisho ngaphandle kwe-GPU acceleration.

I-Multi-Language

Insiza isiNgisi, isiFulentshi, isiSpanishi, isiHindi, isiJapane, isiChinese, isiItalian, isiPutukezi, nesiKorean.

Ukuxhuma umsindo

Uxhume izizwi ezimbili ndawonye ukuze udale ukuhlanganisa kwezizwi okuhlukile.

Sebenzisa izimo

Isikhathi sangempela sokuxhumana nama-bots kanye nabasebenzi ababonakalayo Ukubuka kuqala okuzenzakalelayo kwe-text-to-speech Ukufakwa kwe-Edge kanye nezinhlelo zokusebenza zeselula Uhlelo lokuphatha i-batch oluphezulu

Indlela yokusetshenziswa I-Kokoro

  1. 1

    Ubhalise mahhala noma hlola idemo

    Dala i-akhawunti emahhala engu-TextToSpeechAI ukuze uthole ama-credits angama-200, noma sebenzisa idemo engabhaliswanga ukulalela iKokoro ngokushesha. Izinga elijwayelekile lisho ukuthi iKokoro ibiza kuphela ama-credits angama-10 ngamagama angama-1000.

  2. 2

    Khetha umsindo we-Kokoro

    Vula isiphequluli somsindo bese ukhetha umsindo we-Kokoro kulimi olufunayo (9 oxhaswe, kusuka eNgisini kuya eJapane nakweKorea). Ungasebenzisa futhi ukuhlanganisa umsindo we-Kokoro ukuxuba umsindo amabili ube yi-combination ejwayelekile.

  3. 3

    Faka umbhalo wakho

    Bhala noma cindezela umbhalo ofuna ukuwukhuluma kumhleli. I-Kokoro iphatha iziqephu ezide ngokuphumelelayo ngenxa ye-82M-parameter yayo elula, etholakala eduze kwe-real-time engine.

  4. 4

    Linganisa isivinini bese udala

    Misela isivinini sokudlala ukuze sifane nesimo sakho sokusetshenziswa, bese ucindezela Ukwenza. I-Kokoro inikeza umsindo ngokushesha kunasikhathi sangempela, ngakho-ke umsindo wakho ulungile ngokushesha.

  5. 5

    Layisha phezulu noma sebenzisa i-API

    Layisha ngezansi umsindo oqediwe njenge MP3 noma WAV, noma usebenzise ukukhishwa kwe-automatic nge-TextToSpeechAI REST API ku-api.texttospeechai.com ukuze usebenze nge-batch nesikhathi sangempela.

I-Kokoro I-API

Yenza ulwimi ngokuzenzakalela usebenzisa i-TextToSpeechAI REST API.

curl -X POST "https://api.texttospeechai.com/v1/generate/" \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "I\u002DKokoro inikeza ukukhuluma okujwayelekile ngejubane elimangalisayo nempumelelo.",
    "voice": "en_US-lessac-medium"
  }'

Imibuzo ebuzwa kaningi

I-Kokoro iyimodeli yokubhala-ukukhuluma elula kakhulu enezilungiselelo ezingu-82 million kuphela. Nakuba incane, ikhiqiza ukukhuluma okuzwakalayo okujwayelekile phakathi kwezilimi eziningi ngejubane elifanayo, ngisho ne-CPU.

Yebo, iKokoro igcwele ilayisense le-Apache 2.0 - ikhowudi kanye nesimo sesisindo. Ingasetshenziswa ngokukhululekileyo ezisebenzisweni zebhizinisi ngaphandle kokungabekezeleleki.

I-Kokoro isekela isiNgisi (i-US ne-British), isiFulentshi, isiSpanishi, isiHindi, isiJapane, isiChinese, isiItalian, isiPutukezi, nesiKorean.

I-Kokoro iyinye yezinhlobo ezisheshayo ze-TTS ezikhona. Ikhiqiza amagama ngokushesha kunasikhathi sangempela sokudlala isivinini ngisho ne-CPU, iyenza ibe ngcono kakhulu kuzinhlelo ezixhumene.

Hayi, iKokoro ayixhasi ukuklonywa kwezwi. Isebenzisa i-library yezwi ehlobene nekhono lokuxhuma izwi. Ukuklonywa kwezwi, sebenzisa i-F5-TTS, i-Chatterbox, i-StyleTTS2, i-OpenVoice, noma i-Tortoise.

I-Kokoro ingaxhuma izizwi ezimbili ukuze ikhiqize ukuhlanganisa okuhlukile. Lokhu kukuvumela ukuthi ukhiqize izimo zezwi ezifanele ngaphandle kokuhlanganiswa kwezwi elidala.

Zonke zilula, zilula. I-Kokoro inesakhiwo esimanje futhi ixhasa ukuxhuma umsindo, kanti i-Piper ine-library enkulu yomsindo. Zonke zilungile kuzinhlelo zesikhathi sangempela.

I-Kokoro ifakwe ukuqhuba ku-CPU futhi idinga amacebo aphansi - cishe ama-300MB. Akukho GPU edingekayo, kepha ukukhawulelwa kwe-GPU kuxhaswa ukuqhubekeka okusheshayo.

Yebo. I-Kokoro ikhiqiza amagama ngokushesha kunalokho okudlalwa ku-CPU, nge-latency ephansi kakhulu, ngakho-ke iyinto engcono kakhulu ye-chatbots, abasizamazisi bokukhuluma, nokusakaza okuqhubekayo. Ubukhulu bayo be-82M-parameter bugcina ukusetshenziswa kwememori kuncane, kwenza kube lula ukufinyelela kuvolumu ephezulu ne-edge.

Ukuxhuma umsindo kuvumela ukuthi uxhume umsindo weKokoro kanye nomunye ukuze wenze uxhumano oluhlukile ngezici ezijwayelekile. Ayikho ukuxhuma umsindo ojwayelekile - awukwazi ukubuyisela umuntu othile kusuka kusampula - kodwa inikeza ubukhulu obuningi kunale library yomsindo oqinile. Ungazama ukuxhuma ngokuqondile ku-TextToSpeechAI umhleli.

Zonke zihamba ngokushesha, i-CPU-friendly standard-tier engines ngaphandle kokuhlanganiswa kwezwi. I-Kokoro iyinto encane kakhulu (ingaphezu kuka-300MB) futhi isekela ukuxhuma kwezwi phakathi kwezilimi ezingu-9, ngenkathi i-MeloTTS ibhekene nezilimi eziningi zase-English kanye nesikhathi sangempela sokuphuma kwezilimi eziningi. Khetha i-Kokoro ukuze ubone i-footprint encane kakhulu nokuxhuma; khetha i-MeloTTS uma ufuna izilimi ezithile.

I-Kokoro iyinjini ejwayelekile ye-tier, ebiza ama-credits angama-10 ngamagama angama-1000 - i-tier ephansi ku-TextToSpeechAI. Ama-akhawunti amasha athola ama-credits angama-200 amahhala, ngakho ungazama i-Kokoro ngaphandle kokukhokha. Lokhu kwenza kube yindlela ebiza kakhulu yokukhiqiza amagama asezingeni eliphakeme.

Technical Specs

  • Generation Speed Very Fast
  • Output Quality Good
  • Voice Cloning Not Supported
  • Languages 9
  • GPU VRAM CPU OK
  • Credits/1000 chars 10

Try I-Kokoro Now

Generate your first audio free. No credit card required.

Start Free