Umbhalo usuka kumazwi Ama-AI Agents Umculo Umsindo kuya kumbhalo Ukulungiswa kwezwi Uhlelo lwezwi Imakethe Okuningi

Umthombo ovulekile we-Text to Speech Models

Yonke imodeli ye-TTS kwi-platform yethu ivulekile ngomthombo welayisense efanelekayo yebhizinisi. MIT, Apache 2.0 — akukho ukuvala okusemthethweni, akukho ukuvimbela kokusetshenziswa, akukho zindleko zokugunyazwa okungenzeka. Sebenzisa ngazo nge-API yethu ehostelwe, noma ubeke i-host yakho ku-infrastructure yakho nge-control egcwele.

Umthombo ovulekile Ilayisense le-MIT Apache 2.0 I-self-hosted GitHub
Umhleli we-TTS ogcwele Amadokhumende we-API

Zama manje

0/500
Imahhala neKokoro, Piper, VITS, MeloTTS
Umsindo wakho okhiqizwe uzovela lapha
Ikhiqizwe
Layisha phezulu
Uthanda i-TTS.ai? Ncoma abangane bakho!
Vula umhleli we-TTS ogcwele

Imiphumela ye-TTS yomthombo ovulekile

Kungani amamodeli avulekile abalulekile kumaphrojekthi akho

Zonke izitifiketi ezivulekile

Imodeli ngayinye ku-TTS.ai isebenzisa ilayisense elivulekile elivulekile. Akukho bhokisi elimnyama elisemthethweni, akukho umhlinzeki ovala ngaphakathi, akukho zindleko zokugunyazwa okungalindelekile.

MIT / Apache 2.0

Amamodeli avunyelwe ngaphansi kwe-MIT noma i-Apache 2.0, izinsiza ezivulekile ezivunyelwe kakhulu. Sebenzisa ngokuhweba, guqula, qoqa futhi — akukho kunqande.

I-self-hosted

Layisha ngezansi noma iyiphi imodeli bese uyiqhuba kwihardware yakho. Ukulawula okuphelele kudatha yakho, ukuphuma kwesikhathi, kanye nesakhiwo. Akukho sidingo sokungathembeki kwe-cloud.

GPU engcono kakhulu

Amamodeli alungele i-NVIDIA GPUs ne-CUDA support. I-Piper isebenza ku-CPU kuphela. Amamodeli amaningi adinga i-2-8GB VRAM ukuze kusebenze ukubikezela.

Inhlangano elondolozwe

Imindeni esebenzayo evulekile igcina futhi ithuthukise lezi zinhlobo. Izithonyelwe zidlalwa — thumela amaphutha, ukuthuthukiswa, namazwi amasha ku-GitHub.

Ukusetshenziswa kwebhizinisi OK

Zonke imodeli zivumela ukusetshenziswa kwebhizinisi ngaphansi kwelayisense labo. Yenza imikhiqizo, uthengise izinsizakalo, futhi udale okuqukethwe kwebhizinisi ngaphandle kwe-royalties noma izindleko zokusetshenziswa.

I-Open Source Model Catalog yethu

Yonke imodeli, ilayisense layo, nokuthi isebenza kanjani kahle

KokoroKokoro

Free

Lightweight 82M parameter model delivering studio-quality speech with blazing-fast inference.

Fast 5/5

Okungcono kakhulu: Apache 2.0 — imodeli esezingeni eliphakeme, 82M params, elula ukuphatha ngokwayo

Zama Kokoro

PiperPiper

Free

A fast, local neural text to speech system optimized for Raspberry Pi and embedded devices.

Fast 3/5

Okungcono kakhulu: MIT — CPU kuphela, ilungile kumadivayisi engxenyeni kanye nokuhoxiswa kwe-self-hosting okufakwe

Zama Piper

VITSVITS

Free

Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech.

Fast 3/5

Okungcono kakhulu: MIT — isakhiwo esiyinhloko esisetshenziswa ngamamodeli amaningi aphansi

Zama VITS

BarkBark

Standard

Transformer-based text-to-audio model that generates realistic speech, music, and sound effects.

Slow 4/5

Okungcono kakhulu: MIT — ukukhiqizwa komsindo okuhlukile ngaphezu kwe-TTS ejwayelekile

Zama Bark

Tortoise TTSTortoise TTS

Premium

Multi-voice text-to-speech focused on quality with autoregressive architecture.

Slow 5/5 Ukulungiswa kwezwi

Okungcono kakhulu: Apache 2.0 — ubukhulu bekhwalithi, ukubuyekezwa okubanzi kokucwaninga

Zama Tortoise TTS

OpenVoiceOpenVoice

Premium

Instant voice cloning with granular control over style, emotion, and accent.

Medium 4/5 Ukulungiswa kwezwi

Okungcono kakhulu: MIT — ukuklonya umsindo ovulekile-umthombo ngesimo sokulawula esincane

Zama OpenVoice

Indlela yokusetshenziswa kwe-Open Source TTS

Sebenzisa i-API yethu ehostelwe noma uqhube amamodeli ngokwakho

1

Thola amamodeli avulekile

Thola i-catalog yethu ye-20+ open-source TTS models. Ikhasi ngalinye lemodeli libonisa ilayisense, i-architecture, izimfanelo, kanye nezidingo zokuqasha.

2

Zama kwi-Browser yakho

Ukuhlolwa kwemodeli ngayinye ngokuqondile ku-TTS.ai ngaphandle kokufaka noma yini. Amaseva ethu we-GPU aphatha ukucubungula ukuze ukwazi ukulinganisa ukhwalithi ngaphambi kokufaka i-self-hosting.

3

Usizo lwe-API

Uklonyelisa imodeli ye-repos kusuka ku-GitHub futhi uqhube endaweni, noma sebenzisa i-API yethu ehostelwe ukukhishwa. Ukuhoxiswa kwe-self-hosting kunikeza ukulawula okuphelele; i-API yethu inikeza isakhiwo esilawulwayo.

4

Dala isisebenziso sakho

I-TTS ixhunywe kumkhiqizo wakho usebenzisa amamodeli abekwe ngokwezifiso noma i-REST API yethu. Onke amamodeli angasetshenziswa ngokuhweba ngaphandle kwezindleko zokugunyazwa noma izibopho.

Ukuqhathaniswa kwelayisense

Zonke izimodeli ku-TTS.ai zisebenzisa izilayisense ezivulekile ezilungele ukuthengiswa

Imodeli Ilayisense Ukusetshenziswa kwebhizinisi Ukushintsha Umphathi-we-wedwa Ukunikezwa
Kokoro Apache 2.0 Kudingeka
Piper MIT Okukhethwa kukho
VITS MIT Okukhethwa kukho
MeloTTS MIT Okukhethwa kukho
Chatterbox MIT Okukhethwa kukho
Tortoise TTS Apache 2.0 Kudingeka
StyleTTS 2 MIT Okukhethwa kukho
OpenVoice MIT Okukhethwa kukho
Sesame CSM Apache 2.0 Kudingeka
Orpheus Llama 3.2 "Built with Llama"
Try These Models Free

Ukuhlala ngokwezifiso vs. Ukuhlala API

Sebenzisa amamodeli wena noma sivumele siphathe isakhiwo

Self-Host on Your Hardware

Yonke imodeli ku-TTS.ai itholakala njengephrojekthi yomthombo ovulekile ku-GitHub noma ku-Hugging Face. Layisha ngezansi amasisindo, ufake izi dependances, futhi uqhube ukubikezela ku-GPUs zakho. Unomkhawulo ophelele wokulawula, ubumfihlo, nokukala.

  • Ukuvikelwa kwedatha okuphelele — umsindo awusoze ushiya isisebenzisi sakho
  • Akukho zindleko ezidingayo ngemuva kokumiswa kokuqala
  • Ukuhlela ngokuzimela kudatha yakho
  • Idinga i-hardware ye-GPU (i-NVIDIA ivunyelwe)
  • Uphatha ukuhlaziywa, ukukala, kanye nokwethembeka

Sebenzisa i-TTS.ai Hosted API

Ukuthola ukufinyelela ngokushesha kuzo zonke 20 + imodeli nge REST API eyodwa. Siphatha GPU ukuhlinzekwa, imodeli ukuhlaziywa, ukulawula iqoqo, kanye nokulinganisa. I-API eyodwa inkinobho ikunikeza ukufinyelela kunoma iyiphi imodeli - akukho sidingo sokuphatha izisebenzisi ezahlukene.

  • Akukho mishini ye-GPU edingekayo
  • Zonke imodeli ezingu-20+ nge-API eyodwa
  • Ukuhlaziywa nokuthuthukiswa kwemodeli ngokuzenzakalela
  • 99.9% uptime ngesakhiwo esiningi
  • Imali kuphela ngento oyisebenzisayo
Use Hosted API Instead

Quick Start: API or Self-Host

Sebenzisa i-API yethu ehostelwe, noma ufake i-Kokoro endaweni emizuzwini

Ukhetho 1: TTS.ai Hosted API Elula kakhulu 
import requests

response = requests.post("https://api.tts.ai/v1/tts", json={
    "text": "Open source TTS with a simple API.",
    "model": "kokoro",
    "voice": "af_heart",
    "format": "wav"
}, headers={"Authorization": "Bearer YOUR_API_KEY"})

with open("output.wav", "wb") as f:
    f.write(response.content)
Ukhetho 2: I-self-host nge-pip Ukulawula okuphelele 
# Install Kokoro locally
pip install kokoro

# Generate speech on your own GPU
import kokoro

pipeline = kokoro.KPipeline(lang_code="a")
generator = pipeline("Hello from your own server!", voice="af_heart")
for i, (gs, ps, audio) in enumerate(generator):
    kokoro.save(audio, f"output_{i}.wav")
View API Documentation

Umthombo ovulekile, ukuthengwa okunethezeka

I-API yethu ehostelwe yenza ukuthi i-open-source TTS ifinyeleleke ngaphandle kokuphatha ama-GPUs.

Izinga elikhululekile

$0

15,000 amaphawu ngesikhathi sokubhalisa

  • 4 amamodeli avulekile-source amahhala
  • Akukho ubhaliso lokusetshenziswa okujwayelekile
  • Ukusetshenziswa kwebhizinisi kuvunyelwe

Isiqalisi

$9

500,000 characters/month

  • Zonke imodeli ezingu-20+ ezivulekile
  • Ukuklona umsindo
  • Ukufinyelela kwe-API

I-Pro

$29

2,000,000 characters/month

  • Ukuphathwa kwe-GPU okunesihluthulelo
  • Zonke imodeli eziphezulu
  • Usizo lwebhizinisi
Bona ukuthengiselana okuphelele

Imibuzo ebuzwa kaningi

Imibuzo ejwayelekile mayelana ne-open source text to speech

Yebo. Yonke imodeli ku-TTS.ai isebenzisa ilayisense elivulekile elivumelayo — noma i-MIT noma i-Apache 2.0. Sifaka ngaphandle amamodeli anelayisense avimbelayo (njenge-Coqui's CPML noma i-CC-BY-NC engabizi). Ungaqinisekiswa ilayisense lemodeli ngayinye kwi-GitHub yayo.

Zonke ziyilayisense ezivulekile ezivumela ukusetshenziswa kokuthengiswayo, ukushintshwa, nokuhlukaniswa kabusha. I-Apache 2.0 ifaka izivumelwano ezicacile zepatent futhi idinga ukuchaza izinguquko uma ushintsha ikhowudi. I-MIT ilula kakhulu ngezidingo ezincane. Zonke ziyibhizinisi elilungele.

Yebo. Imodeli ngayinye ingagcinwa ngokuzimela. Khumbula imodeli yendawo yokugcina kusuka ku-GitHub, ufake izithonya, zulazula isisindo semodeli, futhi uqhube ukubikezela. Sinikeza uxhumanisi lwezidingo zokugcinwa kwemodeli ngayinye kufaka phakathi i-GPU, i-RAM, ne-Python version.

Izidingo zihluka ngokwemodeli. I-Piper ayidingi i-GPU (i-CPU kuphela). I-Kokoro ne-MeloTTS zidinga i-1-2GB VRAM. Izinhlobo eziningi ezijwayelekile zidinga i-4GB VRAM. I-Tortoise ne-Sesame CSM zidinga i-8GB. I-NVIDIA RTX 3060 (12GB) ingaqhuba izinhlobo eziningi ngokunethezeka.

Yebo. Amalayisense omthombo ovulekile avumela ukuguqulwa kufaka phakathi ukulungisa okuncane. Amamodeli afana ne-GPT-SoVITS ne-Bark ahlinzeka ngezikripthi zokulungisa okuncane. Ungaqeqesha amamodeli kudatha yomsindo wakho ukuze udale umsindo okhethekile noma uthuthukise ukusebenza kwezilimi ezithile.

Imodeli evulekile ephezulu (iKokoro, iStyleTTS 2, iChatterbox) manje ifana noma idlula izinsizakalo zebhizinisi ezifana ne-ElevenLabs ne-Google TTS ezingeni lomgangatho. Inzuzo enkulu yezinsizakalo zebhizinisi yindawo yokusebenza elawulwayo nexhaso, hhayi umgangatho wesandi.

Sizisuse. XTTS/XTTS-v2 (Coqui's CPML — engabizi), F5-TTS (CC-BY-NC — engabizi), kanye neHiggs-v2 (iBoson License — evimbelayo) zonke zasuswa. Yonke imodeli ku-TTS.ai iqinisekisiwe ukuthi iphephile ekusetshenzisweni kwebhizinisi.

Yebo. Amamodeli amaningi amukela izithobo zeqembu nge-GitHub. Ungathumela izibikezelo zephutha, ukurekhodwa kwezwi ngemibhalo entsha, ukuthuthukiswa kwekhodi, kanye nedokhumende. Khangela i-GitHub yemodeli ngayinye yendawo yokugcina izithobo ngezincomo zokuthobozela kanye nezinkinga ezisebenzayo.

Layisha amamodeli ngokudinga futhi ulayishe lapho ungekho esebenza ukuhlukanisa i-GPU memory. I-GPU server yethu isebenza ngamamodeli angama-20+ ku-4x Tesla P40 (96GB VRAM ephelele) usebenzisa ukulayisha okuqhubekayo. Ukuqasha, i-24GB GPU eyodwa ingasiza amamodeli angama-3-5 ngokufanayo.

Amamodeli amaningi anikeza izithombe zeDocker ezisemthethweni noma amafayela weDocker. Ukusebenza kwamamodeli amaningi, ungakwakha isilungiselelo seDocker esikhethekile nge-NVIDIA Container Toolkit ukungena kwe-GPU. Ukwakhiwa kweseva ye-API yethu kungasetshenziswa njengesicelo sokwethula.

Imodeli eminingi idinga i-Python 3.10-3.12. I-Coqui TTS (VITS) idinga i-Python 3.11. Sicebisa i-Python 3.12 kumodeli eminingi. Khangela i-requirements.txt yemodeli ngayinye ukuze ubone ukuthi iguqulo lihambisana kanjani.

Yebo. I-MIT ne-Apache 2.0 ivumela ukusetshenziswa kokuthengiswayo. Ungakwakha imikhiqizo ye-SaaS, ama-apps eselula, imidlalo, nezinsizakalo usebenzisa lezi zinhlobo ngaphandle kwezindleko zokufaka ilayisense, ama-royalties, noma izimfuneko zokuphawula (noma ngabe ukuphawula kuphawulwe).
5.0/5 (1)

Yini esingayithuthukisa? Umbono wakho usiza ukuxazulula izinkinga.

Zama i-Open Source TTS namhlanje

20 + open-source models, zonke ezisemthethweni-ebhizinisini. Sebenzisa i-API yethu noma i-self-host - ukhetho lukhona.

Ubhalise Bonisa ukuthengwa