AI Text to SpeechName
Konvèti tèks nan pale son natirèl ak 24 + open-source modèl AI. Gratis pou itilize, pa gen okenn kont nesesè.
Wrap ou tèks nan SSML tags pou presizyon kontwòl:
<speak><prosody rate="slow">Slow speech</prosody></speak>
Ajoute marqueurs emotion pou enfliyanse livrezon (modele sipò varye):
Define prononciations Custom (mot = prononciation):
Detay modèl
Dia 2
Dia2 by Nari Labs is a streaming-first upgrade to Dia, available in 1B and 2B parameter variants. It begins synthesizing audio from the first few tokens, making it ideal for real-time voice agents and speech-to-speech pipelines. Supports multi-speaker dialogue with [S1]/[S2] tags and paralinguistic cues like (laughs), (coughs).
| Pwogramè: | Nari Labs |
| Lisans: | Apache 2.0 |
| Vitès | Fast |
| Kalite: | |
| lang | 1 lang |
| VRAM | 4GB |
| Klonaj Vokal | Pa sipòte |
Konsèy pou pi bon rezilta
- Itilize ponktiasyon kòrèk pou pause ak intonasyon natirèl
- Espelye nimewo ak abreviasyon pou yon prononsyasyon pi klè
- Ajoute virgo pou kreye pauze kout ant fraz
- Itilize ellips (...) pou pauze dramatik pi long
- Eseye Kokoro oswa CosyVoice 2 pou rezilta ki pi natirèl
- Itilize Dia pou dyalòg ak kontni podcast
Kout kredi
| Nivo | Koute pou chak 1K karaktè |
|---|---|
| Gratis | 0 kredi (sans limit) |
| Standart | 2 kredi / 1K karaktè |
| Premium | 4 kredi / 1K karaktè |
Kijan AI Text to Speech travay
Pwodui voiceovers pwofesyonèl-kalite nan twa etap senp. Pa gen konesans teknik ki nesesè.
Enter Your Text
Tape, kole, oswa upload tèks ou vle konvèti nan pale. Supports jiska 5, 000 karaktè pa jenerasyon pou itilizatè ki anrejistre. Sèvi ak tèks senp oswa ajoute atik SSML pou kontwòl avanse sou prononciation, pause, ak enfatize.
Chwazi Modèl & Voy
Chwazi soti nan 20 + AI modèl sou twa nivo.Choose yon vwa ki matche ak kontni ou, chwazi lang ou, ajiste vitès playback de 0.5x a 2.0x, epi chwazi fòma ou pi renmen (MP3, WAV, OGG, oswa FLAC).
Kreye & Enstale
Klike sou Pwodui e son ou pral pare nan kèk segonn. Preview ak jwè a enkòpore, telechaje nan fòma ou chwazi a, oswa kopye yon lyen ki ka pataje. Itilize API pou pwosesis batch ak integrasyon nan travay ou.
Konvèti tèks nan pale
Tekst-a-parole ki sipòte pa AI ap chanje fason moun kreye, itilize, ak entèraksyon ak kontni odyo nan dè santèn de endistri.
Tout modèl tèks-a-vokal
Espesifikasyon detaye pou chak modèl AI ki disponib sou TTS.ai.Compare bon jan kalite, vitès, sipò lang, ak karakteristik pou jwenn modèl la ideyal pou pwojè ou.
Kokoro
Free
Kokoro se yon 82 milyon paramèt tèks-a-parole modèl ki punches byen pi wo pase klas pwa li. Pandan ke gwosè li ti, li pwodwi pale remarkabman natirèl ak ekspresif. Kokoro sipòte plizyè lang ki gen ladan angle, Japonè, Chinwa, ak Koreyen ak yon varyete de vwa ekspresif. Li kouri incredibly vit — jenere son prèske 100x pi vit pase tan reyèl sou yon GPU.
Hexgrad
Apache 2.0
Fast
en, ja, zh, ko, fr, de, it, pt, es, hi, ru
1.5GB
Non
Gratis
Piper
Free
Piper se yon motè tèks-a-parole limyè devlope pa Rhasspy ki itilize VITS ak larynx achitekti. Li kouri konplètman sou CPU, ki fè li ideyal pou aparèy edge, automatisation kay, ak aplikasyon ki mande TTS offline. Avèk plis pase 100 vwa nan plis pase 30 lang, Piper bay pale son natirèl nan vitès tan reyèl menm sou yon Raspberry Pi 4.
Rhasspy
MIT
Fast
en, de, fr, es, it, pt, nl, pl, ru, zh, ja, ko, ar, cs, da, fi, el, hu, is, ka, kk, ne, no, ro, sk, sr, sv, sw, tr, uk, vi
0 (CPU only)
Non
Gratis
VITS
Free
VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) se yon metòd TTS paralèl bout-a-bòd ki kreye yon son ki pi natirèl pase modèl aktyèl ki baze sou de etap. Li adopte inférence variational ki ogmante ak koule normalisation ak yon pwosesis antrenman adversarial, rive jwenn yon amelyorasyon siyifikatif nan natiralizasyon.
Jaehyeon Kim et al.
MIT
Fast
en, zh, ja, ko
1GB
Non
Gratis
MeloTTS
Free
MeloTTS by MyShell.ai is a multilingual TTS library supporting English (American, British, Indian, Australian), Spanish, French, Chinese, Japanese, and Korean. It is extremely fast, processing text at near real-time speed on CPU alone. MeloTTS is designed for production use and supports both CPU and GPU inference.
MyShell.ai
MIT
Fast
en, es, fr, zh, ja, ko
0.5GB (GPU optional)
Non
Gratis
Bark
Standard
Bark pa Suno se yon modèl tèks-a-son ki baze sou transformateur ki ka jenere pale trè reyèl, multi-lang kòm byen ke lòt son tankou mizik, bri fon, ak efè son. Li ka pwodwi kominikasyon nonverbal tankou ri, souf, ak plezi. Bark sipòte plis pase 100 preset oratè ak 13 + lang.
Suno
MIT
Slow
en, zh, fr, de, hi, it, ja, ko, pl, pt, ru, es, tr
5GB
Non
2x
Bark Small
Standard
Bark Small se yon vèsyon distilasyon nan modèl Bark ki echanj kèk bon jan kalite son pou vitès infèrans ki pi vit ak kondisyon memwa ki pi ba.Li rete kapasite Bark a pou jenere pale ak emosyon, ri, ak plizyè lang.
Suno
MIT
Medium
en, zh, fr, de, hi, it, ja, ko, pl, pt, ru, es, tr
2GB
Non
2x
CosyVoice 2
Standard
CosyVoice 2 by Alibaba's Tongyi Lab achieves human-comparable speech quality with extremely low latency, making it ideal for real-time applications.It uses a finite scalar quantification approach for streaming synthesis and supports zero-shot voice cloning, cross-lingual synthesis, and fine-grained emotion control.It outperforms many commercial TTS systems in subjective evaluations.
Alibaba (Tongyi Lab)
Apache 2.0
Medium
en, zh, ja, ko, fr, de, it, es
4GB
Wi
2x
Dia TTS
Standard
Dia pa Nari Labs se yon 1.6B paramèt tèks-a-parole modèl ki fèt espesyalman pou jenere dyalòg multi-pale. Li ka pwodwi konvèsasyon son natirèl ant de pale ak approprié turn-taking, prosody, ak ekspresyon emosyonèl. Dia se pafè pou kreye kontni podcast-estil, dyalòg audiobook, ak konvèsasyon entèaktif AI.
Nari Labs
Apache 2.0
Medium
en
4GB
Non
2x
Parler TTS
Standard
Parler TTS se yon modèl tèks-a-vokal ki sèvi ak dekriman vwa lang natif natal pou kontwole vwa ki pwodwi a. An plas pou chwazi soti nan vwa presegondè, ou dekri vwa ou vle a (e. g., "yon vwa fanm cho ak yon ti aksan Britanik, ki pale byen vit ak klèman") epi Parler jenere vwa ki sanble ak dekriman sa a. Sa fè li inikman fleksib pou aplikasyon kreyatif.
Hugging Face
Apache 2.0
Medium
en
4GB
Non
2x
GLM-TTS
Standard
GLM-TTS pa Zhipu AI se yon sistèm tèks-a-parole ki bati sou achitekti Llama ak matche ak koule. Li rive nan pi ba pousantaj erè karaktè nan mitan modèl TTS ki gen sous louvri, sa vle di li pwodwi prononsyasyon ki pi egzat. GLM-TTS sipòte angle ak Chinwa ak klonaj vwa soti nan 3-10 dezyèm echantiyon son.
Zhipu AI
GLM-4 License
Medium
en, zh
4GB
Wi
2x
IndexTTS-2
Standard
IndexTTS-2 se yon sistèm avanse tèks-a-parole ki ekselan nan sintèz vwa zero-shot ak kontwòl emosyon fine-grained. Li ka jenere pale ak ton emosyonèl espesifik tankou kè kontan, trist, kole, oswa terib san yo pa mande done fòmasyon emosyonèl espesifik. Modèl la itilize vektè emosyonèl pou kontwole egzakteman ekspresyon emosyonèl nan pale jenere.
Index Team
Bilibili Model License
Medium
en, zh
4GB
Wi
2x
Spark TTS
Standard
Spark TTS pa SparkAudio se yon modèl tèks-a-parole ki konbine klonaj vwa ak efè ki ka kontwole ak style pale. Avèk jis 5 segonn nan son referans, li ka klone yon vwa epi Lè sa a, jenere pale ak diferan efè, vitès, ak style pandan y ap kenbe idantite vwa klone. Spark TTS itilize yon sistèm kontwòl ki baze sou pwompt.
SparkAudio
CC BY-NC-SA 4.0
Medium
en, zh
4GB
Wi
2x
GPT-SoVITS
Standard
GPT-SoVITS konbine modélisation langaj GPT-style ak SoVITS (Singing Voice Inference via Translation and Synthesis) pou yon klonaj voyifikatif. Avèk sèlman 5 segonn son referans, li kapab klone yon voyi ak presizyon epi jenere yon nouvo pale pandan l ap kenbe karakteristik unike pale a.
RVC-Boss
MIT
Slow
en, zh, ja, ko
6GB
Wi
2x
Orpheus
Standard
Orpheus se yon modèl tèks-a-parole gwo-echèl ki rive jwenn ekspresyon emosyonèl nan nivo moun. Trete sou plis pase 100,000 èdtan nan done pale divès, li ekselans nan jenerasyon pale ak emosyon natirèl, enpak, ak style pale. Orpheus ka pwodwi pale ki se prèske indistinguishable de enregistrements moun.
Canopy Labs
Llama 3.2 Community
Medium
en
4GB
Non
2x
Chatterbox
Premium
Chatterbox by Resemble AI se yon modèl klonaj vwa ki pi avanse. Li kapab replike nenpòt vwa soti nan yon echantiyon odyo sèl ak yon presizyon resanblans, kaptive pa sèlman timbre men tou style pale ak nuans emosyonèl. Chatterbox gen tou kontwòl emosyonèl fin-grained, ki pèmèt ou ajiste ton emosyonèl nan pale jenere endepandan de idantite vwa.
Resemble AI
MIT
Medium
en
4GB
Wi
4x
Tortoise TTS
Premium
Tortoise TTS se yon sistèm tèks-a-parole multi-vokal autoregressif ki bay priyorite a bon jan kalite odyo sou vitès. Li itilize DALL-E-inspired achitekti pou jenere pale trè natirèl ak ekselan prosody ak pale similitude. Pandan ke pi lejè pase anpil altènativ, Tortoise pwodwi kèk nan pale sintèz ki pi reyèl disponib nan open-source ekosistem.
James Betker
Apache 2.0
Slow
en
8GB
Wi
4x
StyleTTS 2
Premium
StyleTTS 2 reyalize TTS sentez nivo moun pa konbine difizyon estil ak antrenè kontwovèsyal ki itilize gwo modèl lang pale. Li jenere pale sonje ki pi natirèl ant modèl yon sèl-pale, rival enregistrements moun. StyleTTS 2 itilize difizyon-ki baze sou modélisation estil pou capture la gamme complète de varyasyon pale moun.
Columbia University
MIT
Medium
en
4GB
Non
4x
OpenVoice
Premium
OpenVoice pa MyShell.ai pèmèt klonaj vwa imedyat ak kontwòl granulaire sou style vwa, emosyon, aksan, rythme, pause, ak intonasyon. Li ka klone yon vwa soti nan yon clip son kout ak jenere pale nan plizyè lang pandan y ap kenbe idantite pale a. OpenVoice tou fonksyone kòm yon konvètisè vwa, ki pèmèt transformasyon vwa nan tan reyèl.
MyShell.ai / MIT
MIT
Medium
en, zh, ja, ko, fr, de, es, it
4GB
Wi
4x
Qwen3 TTS
Standard
Qwen3-TTS se yon modèl 1.7 milya dola paramèt tèks-a-parole ki soti nan ekip Qwen Alibaba a. Li sipòte twa mòd: vwa preset ak kontwòl emosyon (9 oratè), klonaj vwa soti nan jis 3 segonn odyo, ak yon mòd konsepsyon vwa inik kote ou dekri vwa ou vle nan lang natirèl.
Alibaba (Qwen)
Apache 2.0
Medium
en, zh, ja, ko, de, fr, ru, pt, es, it
7GB
Wi
2x
Sesame CSM
Premium
Sesame CSM (Conversational Speech Model) se yon modèl 1 milya dola paramèt ki fèt espesyalman pou jenere pale konvèsasyon. Li modènize modèl natirèl konvèsasyon imen an ki gen ladan tan pran vire, repons backchannel, reaksyon emosyonèl, ak koule konvèsasyon. CSM jenere son ki son tankou yon konvèsasyon imen natirèl plis pase pale sintètik.
Sesame
Apache 2.0
Slow
en
8GB
Non
4x
Chatterbox Turbo
Standard
Chatterbox Turbo by Resemble AI se yon ajou paramèt 350M pou Chatterbox, ki bay jiska 6x vitès tan reyèl ak latency sub-200ms. Li sipòte atik paralingwistik tankou [laugh], [cough], ak [chuckle] dirèkteman nan tèks. Gen ladann Perth watermarking sou tout odyo jenere pou suivi provenance.
Resemble AI
MIT
Fast
en
2GB
Wi
2x
Zonos
Standard
Zonos v0.1 pa Zyphra se yon modèl paramèt 1.6B ki gen ladan kontwòl emosyon fine-grained ak sliders pou kè kontan, kòlè, tristès, rèv, ak surpris. Li ofri tou de yon Transformateur ak yon nouvo SSM (modèl eta-espasyal) varyasyon. Antrene sou 200K + èdtan nan pale multilingual ak zero-shot klonaj vwa soti nan 10-30 segonn nan referans son.
Zyphra
Apache 2.0
Medium
en, ja, zh, fr, de
6GB
Wi
2x
Dia 2
Standard
Dia2 by Nari Labs is a streaming-first upgrade to Dia, available in 1B and 2B parameter variants. It begins synthesizing audio from the first few tokens, making it ideal for real-time voice agents and speech-to-speech pipelines. Supports multi-speaker dialogue with [S1]/[S2] tags and paralinguistic cues like (laughs), (coughs).
Nari Labs
Apache 2.0
Fast
en
4GB
Non
2x
VoxCPM
Standard
VoxCPM 1. 5 pa OpenBMB se yon nouvo modèl TTS san tokenizer ki opere nan espas kontinyèl plis ke tokens diskrèt. Li pwodwi 44. 1kHz son segondè- fidélité, sipòte klonaj vwa zero- shot soti nan 3- 10 segonn, ak kenbe konstans ant paragraf. Cross- lang klonaj pèmèt ou aplike yon vwa angle nan pale Chinwa ak vice versa.
OpenBMB
Apache 2.0
Fast
en, zh
4GB
Wi
2x
OuteTTS
Free
OuteTTS pwolonje gwo modèl lang ak kapasite tèks-a-parole pandan l ap kenbe achitekti orijinal la. Li sipòte backends multiple ki gen ladan llama.cpp (CPU / GPU), Hugging Face Transformers, ExLlamaV2, VLLM, ak menm infèrans navigatè via Transformers.js. Features zero-shot voice cloning through speaker profiles saved as JSON.
OuteAI
Apache 2.0
Fast
en
2GB
Wi
Gratis
TADA
Standard
TADA (Text-Acoustic Dual Alignment) pa Hume AI se yon modèl TTS inovatè ki elimine alucination a travers yon achitekti alignement doub nouvo bati sou Llama 3.2. Disponib nan 1B (an Anglè) ak 3B (multilingual) varyete, TADA rive yon RTF nan 0.09 — 5x pi vit pase modèl TTS ki baze sou LLM konparab.
Hume AI
MIT
Fast
en
5GB
Non
2x
VibeVoice
Standard
VibeVoice by Microsoft vini nan de varyasyon: yon modèl 1.5B pou kontni long-fòm (jouk a 90 minit, 4 oratè) ak yon Realtime 0.5B modèl pou streaming ak ~ 200ms premye latency son. varyasyon 1.5B excels nan podcasts ak liv son ak konstanson oratè sou long pasaj.
Microsoft
MIT
Fast
en, zh
4GB
Non
2x
Pocket TTS
Free
Pocket TTS pa Kyutai (kreyatè Moshi) se yon modèl tèks-a-parole 100M paramèt ki koube byen pi wo pase pwa li. Li kouri efikasman sou CPU, sipòte klonaj vwa zero-shot soti nan yon sèl echantiyon son, epi pwodwi pale natirèl-son. Ti gwosè modèl la fè li ideyal pou deployment edge ak anviwònman ki ba-resous.
Kyutai
MIT
Fast
en, fr
1GB
Wi
Gratis
Kitten TTS
Free
Kitten TTS by KittenML is an ultra-lightweight text-to-speech model built on ONNX. With variants from 15M to 80M parameters (25-80 MB on disk), it delivers high-quality voice synthesis on CPU without requiring a GPU. Features 8 built-in voices, adjustable speech speed, and built-in text preprocessing for numbers, currencies, and units. Ideal for edge deployment and low-latency applications.
KittenML
Apache 2.0
Fast
en
0GB
Non
Gratis
CosyVoice3
Standard
CosyVoice3 is the latest evolution from Alibaba's FunAudioLLM team. It features bi-streaming inference with ~150ms latency, instruction-based control for emotion/speed/volume, and improved speaker similarity for zero-shot cloning. Supports 9 languages plus 18 Chinese dialects. RL-tuned variant delivers state-of-the-art prosody.
Alibaba (FunAudioLLM)
Apache 2.0
Fast
en, zh, ja, ko, de, es, fr, it, ru
4GB
Wi
2x
MOSS-TTS
Premium
MOSS-TTS from OpenMOSS supports generation of up to 1 hour of continuous speech across 20 languages. Features token-level duration control, phoneme-level pronunciation control via IPA/Pinyin, and code-switching between languages. The 8B production model delivers state-of-the-art quality with zero-shot voice cloning from reference audio.
OpenMOSS
Apache 2.0
Medium
en, zh, de, es, fr, ja, it, hu, ko, ru, fa, ar, pl, pt, cs, da, sv, el, tr
16GB
Wi
4x
MegaTTS3
Premium
MegaTTS3 from ByteDance uses a novel sparse alignment mechanism combined with a latent diffusion transformer. Features adjustable trade-off between speech intelligibility and speaker similarity for zero-shot voice cloning.
ByteDance
Apache 2.0
Slow
en, zh
8GB
Wi
4x
Kokoro
Gratis
Kokoro is an 82 million parameter text-to-speech model that punches well above its weight class. Despite its tiny size, it produces remarkably natural and expressive speech. Kokoro supports multiple languages including English, Japanese, Chinese, and Korean with a variety of expressive voices. It runs incredibly fast — generating audio nearly 100x faster than real-time on a GPU.
Hexgrad
Apache 2.0
Fast
Piper
Gratis
Piper is a lightweight text-to-speech engine developed by Rhasspy that uses VITS and larynx architectures. It runs entirely on CPU, making it ideal for edge devices, home automation, and applications requiring offline TTS. With over 100 voices across 30+ languages, Piper delivers natural-sounding speech at real-time speeds even on a Raspberry Pi 4.
Rhasspy
MIT
Fast
VITS
Gratis
VITS (Variational Inference with adversarial learning for end-to-end Text-to-Speech) is a parallel end-to-end TTS method that generates more natural sounding audio than current two-stage models. It adopts variational inference augmented with normalizing flows and an adversarial training process, achieving a significant improvement in naturalness.
Jaehyeon Kim et al.
MIT
Fast
MeloTTS
Gratis
MeloTTS by MyShell.ai is a multilingual TTS library supporting English (American, British, Indian, Australian), Spanish, French, Chinese, Japanese, and Korean. It is extremely fast, processing text at near real-time speed on CPU alone. MeloTTS is designed for production use and supports both CPU and GPU inference.
MyShell.ai
MIT
Fast
OuteTTS
Gratis
OuteTTS extends large language models with text-to-speech capabilities while preserving the original architecture. It supports multiple backends including llama.cpp (CPU/GPU), Hugging Face Transformers, ExLlamaV2, VLLM, and even browser inference via Transformers.js. Features zero-shot voice cloning through speaker profiles saved as JSON.
OuteAI
Apache 2.0
Fast
Pocket TTS
Gratis
Pocket TTS by Kyutai (creators of Moshi) is a compact 100M parameter text-to-speech model that punches well above its weight. It runs efficiently on CPU, supports zero-shot voice cloning from a single audio sample, and produces natural-sounding speech. The small model size makes it ideal for edge deployment and low-resource environments.
Kyutai
MIT
Fast
Kitten TTS
Gratis
Kitten TTS by KittenML is an ultra-lightweight text-to-speech model built on ONNX. With variants from 15M to 80M parameters (25-80 MB on disk), it delivers high-quality voice synthesis on CPU without requiring a GPU. Features 8 built-in voices, adjustable speech speed, and built-in text preprocessing for numbers, currencies, and units. Ideal for edge deployment and low-latency applications.
KittenML
Apache 2.0
Fast
Bark
Standart
Bark by Suno is a transformer-based text-to-audio model that can generate highly realistic, multilingual speech as well as other audio like music, background noise, and sound effects. It can produce nonverbal communications like laughing, sighing, and crying. Bark supports over 100 speaker presets and 13+ languages.
Suno
MIT
Slow
en, zh, fr, de, hi, it, ja, ko, pl, pt, ru, es, tr
Non
Bark Small
Standart
Bark Small is a distilled version of the Bark model that trades some audio quality for significantly faster inference speeds and lower memory requirements. It retains Bark's ability to generate speech with emotions, laughter, and multiple languages.
Suno
MIT
Medium
en, zh, fr, de, hi, it, ja, ko, pl, pt, ru, es, tr
Non
CosyVoice 2
Standart
CosyVoice 2 by Alibaba's Tongyi Lab achieves human-comparable speech quality with extremely low latency, making it ideal for real-time applications. It uses a finite scalar quantization approach for streaming synthesis and supports zero-shot voice cloning, cross-lingual synthesis, and fine-grained emotion control. It outperforms many commercial TTS systems in subjective evaluations.
Alibaba (Tongyi Lab)
Apache 2.0
Medium
en, zh, ja, ko, fr, de, it, es
Wi
Dia TTS
Standart
Dia by Nari Labs is a 1.6B parameter text-to-speech model designed specifically for generating multi-speaker dialogue. It can produce natural-sounding conversations between two speakers with appropriate turn-taking, prosody, and emotional expression. Dia is perfect for creating podcast-style content, audiobook dialogues, and interactive conversational AI.
Nari Labs
Apache 2.0
Medium
en
Non
Parler TTS
Standart
Parler TTS is a text-to-speech model that uses natural language voice descriptions to control the generated speech. Instead of selecting from preset voices, you describe the voice you want (e.g., "a warm female voice with a slight British accent, speaking slowly and clearly") and Parler generates speech matching that description. This makes it uniquely flexible for creative applications.
Hugging Face
Apache 2.0
Medium
en
Non
GLM-TTS
Standart
GLM-TTS by Zhipu AI is a text-to-speech system built on the Llama architecture with flow matching. It achieves the lowest character error rate among open-source TTS models, meaning it produces the most accurate pronunciation. GLM-TTS supports English and Chinese with voice cloning from 3-10 second audio samples.
Zhipu AI
GLM-4 License
Medium
en, zh
Wi
IndexTTS-2
Standart
IndexTTS-2 is an advanced text-to-speech system that excels at zero-shot voice synthesis with fine-grained emotion control. It can generate speech with specific emotional tones like happy, sad, angry, or fearful without requiring emotion-specific training data. The model uses emotion vectors to precisely control the emotional expression of generated speech.
Index Team
Bilibili Model License
Medium
en, zh
Wi
Spark TTS
Standart
Spark TTS by SparkAudio is a text-to-speech model that combines voice cloning with controllable emotion and speaking style. Using just 5 seconds of reference audio, it can clone a voice and then generate speech with different emotions, speeds, and styles while maintaining the cloned voice identity. Spark TTS uses a prompt-based control system.
SparkAudio
CC BY-NC-SA 4.0
Medium
en, zh
Wi
GPT-SoVITS
Standart
GPT-SoVITS combines GPT-style language modeling with SoVITS (Singing Voice Inference via Translation and Synthesis) for powerful few-shot voice cloning. With as little as 5 seconds of reference audio, it can accurately clone a voice and generate new speech while preserving the speaker's unique characteristics. It excels at both speaking and singing voice synthesis.
RVC-Boss
MIT
Slow
en, zh, ja, ko
Wi
Orpheus
Standart
Orpheus is a large-scale text-to-speech model that achieves human-level emotional expression. Trained on over 100,000 hours of diverse speech data, it excels at generating speech with natural emotions, emphasis, and speaking styles. Orpheus can produce speech that is virtually indistinguishable from human recordings.
Canopy Labs
Llama 3.2 Community
Medium
en
Non
Qwen3 TTS
Standart
Qwen3-TTS is a 1.7 billion parameter text-to-speech model from Alibaba's Qwen team. It supports three modes: preset voices with emotion control (9 speakers), voice cloning from just 3 seconds of audio, and a unique voice design mode where you describe the voice you want in natural language. It covers 10 languages with high expressiveness and natural prosody.
Alibaba (Qwen)
Apache 2.0
Medium
en, zh, ja, ko, de, fr, ru, pt, es, it
Wi
Chatterbox Turbo
Standart
Chatterbox Turbo by Resemble AI is a 350M parameter upgrade to Chatterbox, delivering up to 6x real-time speed with sub-200ms latency. It supports paralinguistic tags like [laugh], [cough], and [chuckle] directly in text. Includes Perth watermarking on all generated audio for provenance tracking.
Resemble AI
MIT
Fast
en
Wi
Zonos
Standart
Zonos v0.1 by Zyphra is a 1.6B parameter model featuring fine-grained emotion control with sliders for happiness, anger, sadness, fear, and surprise. It offers both a Transformer and a novel SSM (state-space model) variant. Trained on 200K+ hours of multilingual speech with zero-shot voice cloning from 10-30 seconds of reference audio.
Zyphra
Apache 2.0
Medium
en, ja, zh, fr, de
Wi
Dia 2
Standart
Dia2 by Nari Labs is a streaming-first upgrade to Dia, available in 1B and 2B parameter variants. It begins synthesizing audio from the first few tokens, making it ideal for real-time voice agents and speech-to-speech pipelines. Supports multi-speaker dialogue with [S1]/[S2] tags and paralinguistic cues like (laughs), (coughs).
Nari Labs
Apache 2.0
Fast
en
Non
VoxCPM
Standart
VoxCPM 1.5 by OpenBMB is a novel tokenizer-free TTS model that operates in continuous space rather than discrete tokens. It produces high-fidelity 44.1kHz audio, supports zero-shot voice cloning from 3-10 seconds, and maintains consistency across paragraphs. Cross-language cloning lets you apply an English voice to Chinese speech and vice versa.
OpenBMB
Apache 2.0
Fast
en, zh
Wi
TADA
Standart
TADA (Text-Acoustic Dual Alignment) by Hume AI is a groundbreaking TTS model that eliminates hallucinations through a novel dual alignment architecture built on Llama 3.2. Available in 1B (English) and 3B (multilingual) variants, TADA achieves an RTF of 0.09 — 5x faster than comparable LLM-based TTS models. It supports up to 700 seconds of audio context and produces emotionally expressive speech with zero hallucinations on standard benchmarks.
Hume AI
MIT
Fast
en
Non
VibeVoice
Standart
VibeVoice from Microsoft generates long-form speech up to 90 minutes with support for 4 simultaneous speakers, making it ideal for podcasts and dialogues. The Realtime 0.5B variant achieves ~300ms latency for interactive use. Supports speaker tags for multi-turn dialogue generation.
Microsoft
MIT
Fast
en, zh
Non
CosyVoice3
Standart
CosyVoice3 is the latest evolution from Alibaba's FunAudioLLM team. It features bi-streaming inference with ~150ms latency, instruction-based control for emotion/speed/volume, and improved speaker similarity for zero-shot cloning. Supports 9 languages plus 18 Chinese dialects. RL-tuned variant delivers state-of-the-art prosody.
Alibaba (FunAudioLLM)
Apache 2.0
Fast
en, zh, ja, ko, de, es, fr, it, ru
Wi
Model Tab Komparasyon
| Modèl | Pwogramè: | Nivo | Kalite: | Vitès | lang | Klonaj Vokal | VRAM | Lisans: | Kredi | |
|---|---|---|---|---|---|---|---|---|---|---|
| Kokoro | Hexgrad | Free | Fast | 11 | 1.5GB | Apache 2.0 | Gratis | Itilize | ||
| Piper | Rhasspy | Free | Fast | 31 | 0 (CPU only) | MIT | Gratis | Itilize | ||
| VITS | Jaehyeon Kim et al. | Free | Fast | 4 | 1GB | MIT | Gratis | Itilize | ||
| MeloTTS | MyShell.ai | Free | Fast | 6 | 0.5GB (GPU optional) | MIT | Gratis | Itilize | ||
| Bark | Suno | Standard | Slow | 13 | 5GB | MIT | 2 | Itilize | ||
| Bark Small | Suno | Standard | Medium | 13 | 2GB | MIT | 2 | Itilize | ||
| CosyVoice 2 | Alibaba (Tongyi Lab) | Standard | Medium | 8 | 4GB | Apache 2.0 | 2 | Itilize | ||
| Dia TTS | Nari Labs | Standard | Medium | 1 | 4GB | Apache 2.0 | 2 | Itilize | ||
| Parler TTS | Hugging Face | Standard | Medium | 1 | 4GB | Apache 2.0 | 2 | Itilize | ||
| GLM-TTS | Zhipu AI | Standard | Medium | 2 | 4GB | GLM-4 License | 2 | Itilize | ||
| IndexTTS-2 | Index Team | Standard | Medium | 2 | 4GB | Bilibili Model License | 2 | Itilize | ||
| Spark TTS | SparkAudio | Standard | Medium | 2 | 4GB | CC BY-NC-SA 4.0 | 2 | Itilize | ||
| GPT-SoVITS | RVC-Boss | Standard | Slow | 4 | 6GB | MIT | 2 | Itilize | ||
| Orpheus | Canopy Labs | Standard | Medium | 1 | 4GB | Llama 3.2 Community | 2 | Itilize | ||
| Chatterbox | Resemble AI | Premium | Medium | 1 | 4GB | MIT | 4 | Itilize | ||
| Tortoise TTS | James Betker | Premium | Slow | 1 | 8GB | Apache 2.0 | 4 | Itilize | ||
| StyleTTS 2 | Columbia University | Premium | Medium | 1 | 4GB | MIT | 4 | Itilize | ||
| OpenVoice | MyShell.ai / MIT | Premium | Medium | 8 | 4GB | MIT | 4 | Itilize | ||
| Qwen3 TTS | Alibaba (Qwen) | Standard | Medium | 10 | 7GB | Apache 2.0 | 2 | Itilize | ||
| Sesame CSM | Sesame | Premium | Slow | 1 | 8GB | Apache 2.0 | 4 | Itilize | ||
| Chatterbox Turbo | Resemble AI | Standard | Fast | 1 | 2GB | MIT | 2 | Itilize | ||
| Zonos | Zyphra | Standard | Medium | 5 | 6GB | Apache 2.0 | 2 | Itilize | ||
| Dia 2 | Nari Labs | Standard | Fast | 1 | 4GB | Apache 2.0 | 2 | Itilize | ||
| VoxCPM | OpenBMB | Standard | Fast | 2 | 4GB | Apache 2.0 | 2 | Itilize | ||
| OuteTTS | OuteAI | Free | Fast | 1 | 2GB | Apache 2.0 | Gratis | Itilize | ||
| TADA | Hume AI | Standard | Fast | 1 | 5GB | MIT | 2 | Itilize | ||
| VibeVoice | Microsoft | Standard | Fast | 2 | 4GB | MIT | 2 | Itilize | ||
| Pocket TTS | Kyutai | Free | Fast | 2 | 1GB | MIT | Gratis | Itilize | ||
| Kitten TTS | KittenML | Free | Fast | 1 | 0GB | Apache 2.0 | Gratis | Itilize | ||
| CosyVoice3 | Alibaba (FunAudioLLM) | Standard | Fast | 9 | 4GB | Apache 2.0 | 2 | Itilize | ||
| MOSS-TTS | OpenMOSS | Premium | Medium | 19 | 16GB | Apache 2.0 | 4 | Itilize | ||
| MegaTTS3 | ByteDance | Premium | Slow | 2 | 8GB | Apache 2.0 | 4 | Itilize |
Pi gwo platfòm AI Text to Speech
Poukisa chwazi TTS.ai pou Text to Speech?
TTS.ai pote ansanm mond lan
Chak modèl se sous louvri anba MIT, Apache 2.0, oswa lisans permissive similaire, asire w ke ou gen tout dwa komèsyal pou itilize son jenere nan pwojè ou yo.Si ou bezwen rapid, sintezis limyè pou aplikasyon an tan reyèl oswa ekstraksyon kalite studio-premium pou audiobooks ak podcasts, TTS.ai gen modèl la dwa pou chak ka itilize.
Modèl gratis, pa gen kont mande
Kòmanse imedyatman ak twa modèl TTS gratis: Piper (ultra-vitès, limyè), VITS (sintez nève bon jan kalite segondè), ak MeloTTS (sipò pou plizyè lang). Pa gen enskripsyon, pa gen kat kredi, pa gen limit sou jenerasyon. Modèl gratis yo sipòte angle ak plizyè lòt lang ak rezilta son natirèl ki apwopriye pou pifò aplikasyon.
Pwosesis akselere pa GPU
Tout modèl TTS yo kouri sou GPU NVIDIA dedikatè pou tan jenerasyon vit ak konstan. Modèl gratis yo jeneralman jenere odyo nan mwens pase 2 segonn. Modèl estanda tankou Kokoro, CosyVoice 2, ak Bark gen yon mwayèn de 3-5 segonn. Modèl Premium ak pi bon kalite a, tankou Tortoise ak Chatterbox, trete nan 5-15 segonn depann de longè tèks la.
30+ lang sipòte
Kreye pale nan plis pase 30 lang ki gen ladan angle, Espay, franse, Alman, Italyen, Pòtigè, Chinwa, Japonè, Koreyen, Arab, Hindi, Ris, ak anpil lòt. Diferan modèl sipòte sintezis lang, sa vle di ou ka kreye pale nan yon lang ke vwa orijinal la pa janm te antrene sou. CosyVoice 2 ak GPT-SoVITS ekselan nan klonaj vwa lang.
Developer-Ready API
Enkòpore TTS.ai nan aplikasyon ou yo ak OpenAI-kompatib REST API nou an. Yon pwen fen pou tout 20 + modèl. Python, JavaScript, cURL, ak Go SDKs. Streaming sipò pou aplikasyon an tan reyèl. Pwosesis batch pou jenerasyon kontni gwo-echèl. Webhooks pou notifikasyon async. Disponib sou Pro ak Enterprise plan.
Kesyon ki poze souvan
What could we improve? Your feedback helps us fix issues.
Kòmanse konvèti tèks nan pale kounye a
Join milye de kreyatè ki itilize TTS.ai. Obtenn 15,000 karaktè gratis ak yon nouvo kont. Modèles gratis ki disponib san yo pa enskripsyon.