AI Te kupu ki te kōrero
Ka tahuri te kupu ki roto i te kōrero māori me ngā tauira pūtake tūwhera o AI. Whai te whakamahi, kāore he tatau e hiahiatia ana.
Whāriki i tōna kupu i roto i ngā tohu SSML mō te whakahaere tika:
<speak><prosody rate="slow">Slow speech</prosody></speak>
Tāpiri i ngā tohu ā-āhuatanga hei whakaawe i te tuku (he rerekē te tautoko tauira):
Ka tautuhia ngā tohutohu ā-ringa (wāhi = tohutohu):
Whakamāramatanga tauira
CosyVoice3
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.
| kaiwhakawhanake: | Alibaba (FunAudioLLM) |
| Whakawhiwhinga: | Apache 2.0 |
| Āhuatanga | Fast |
| Kāwai: | |
| reo | 9 reo |
| VRAM | 4GB |
| Ko te tāruatanga reo | Kua tautokona |
Ko ngā tohu mō ngā hua pai ake
- Ka whakamahia te whakawāteatanga tika mō ngā whakawāteatanga māori me ngā whakawāteatanga.
- E whakamāori ana i ngā tau me ngā whakawhāititanga mō te kōrero mārama ake.
- E tāpiri ana i ngā kōwae hei waihanga i ngā wā pōturi i waenganui i ngā rerenga
- Ka whakamahia ngā kōaro (...) mō ngā wā roa ake
- Whakamātau i te Kokoro, i te CosyVoice 2 rānei mō ngā hua tino māori
- Ka whakamahia a Dia mō te tauwhitinga kaikōrero-maha me ngā ihirangi podcast
Ko ngā utu pūtea
| Te āhua | Ko te utu mō ia pūāhua 1K |
|---|---|
| Waihoki | 0 ngā pūtea (kore te tepe) |
| Paerewa | 2 ngā pūtea / 1K ngā pūāhua |
| Whakawhiwhinga | 4 ngā pūtea / 1K ngā pūāhua |
He pēhea te mahi a AI Text-to-Speech
E toru ngā hipanga ngāwari hei waihanga i ngā kōrero ā-mahi. Kāore he mōhiotanga hangarau e hiahiatia ana.
Ka tāuru i ōna kupu
Type, paste, whakaata rānei i te kupu e hiahiatia ana e koe kia tahuri ki te kōrero. E tautoko ana ki te 5,000 ngā tohu i ia whakatupuranga mō ngā kaiwhakaari tāurunga. Ka whakamahia te kupu pūnoa, ka tāpiri rānei i ngā tohu SSML mō te whakahaere matatini i runga i te kōrero, i ngā whakapeka, me ngā whakahua.
Hiko te tauira me te reo
Ka kōwhiria mai i ngā tauira AI 20+ puta noa i ngā taumata e toru. Ka kōwhiria he reo e ōrite ana ki tōna ihirangi, e kōwhiria ana i tōna reo ūnga, e whakarerekē ana i te tere tākaro mai i te 0.5x ki te 2.0x, me te kōwhiri i tōna āhua huaputa e manakohia ana (MP3, WAV, OGG, FLAC rānei).
Ka whakaputaina me te tangohia
Tirohia me te kaiwhakaari whāiti, tuku i roto i tōna hanga e kōwhiria ana, tārua rānei i tētahi pātahitanga tiritiri. Ka whakamahia te API mō te tukanga rōpū me te whakaurutanga ki roto i tōna rerenga mahi.
Ka whakamahia te kupu ki te kōrero
Ko te kupu-ki-whakaahua AI e huri ana i te āhua o te waihanga, te whakapaunga, me te tauwhitinga a te tangata ki ngā ihirangi oro i roto i ngā mahi maha.
Ko ngā tauira kupu ki te kōrero katoa
Ko ngā whakaritenga mōhiohio mō ia tauira AI e wātea ana i TTS.ai. Tērā te āhuatanga, te tere, te tautoko reo, me ngā āhuatanga hei kimi i te tauira tika mō tōmu kaupapa.
Kokoro
Free
Ko te Kokoro he tauira tuhi-ki-te-kōrero tauine 82 miriona e ātete ana i runga ake i tōna karaehe taumaha. Ahakoa tōna rahi iti, ka whakaputaina e ia he kōrero tino māori me te whakamārama. Ko te Kokoro e tautoko ana i ngā reo maha tae atu ki te reo Ingarihi, te reo Hapanihi, te reo Hainamana, me te reo Korean me ngā reo whakamārama maha. He tere rawa — e whakaputa ai i te oro tata ki te 100x tere ake i te wā tūturu i runga i te GPU.
Hexgrad
Apache 2.0
Fast
en, ja, zh, ko, fr, de, it, pt, es, hi, ru
1.5GB
Kāore
Waihoki
Piper
Free
Ko Piper he mīhini kupu-ki-whakaahua ngāwari i hangaia e Rhasspy e whakamahi ana i ngā hanganga VITS me te larynx. E mahi ana katoa ana i runga i te CPU, e pai ana mō ngā pūrere pae, ngā pūkaha kāinga, me ngā taupānga e hiahiatia ana he TTS kāore i te tīariari. Me ngā reo neke atu i te 100 puta noa i ngā reo 30+, e tuku ana a Piper i te kōrero māori i te tere o te wā tūturu i runga anō i te 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)
Kāore
Waihoki
VITS
Free
VITS (He whakarerekētanga me te akoranga ātete mō te mutunga-ki-te mutunga o te kupu-ki-te-whakahaere) he aratuka TTS mutunga-ki-te mutunga e puta ai he pūoro māori ake i ngā tauira wāhanga-rua o nāianei, e whakaae ana ki te whakarerekētanga o te whakarerekētanga i whakanuia e ngā rerenga pūnoa me tētahi tukanga whakaakoranga ātete, e whiwhi ana i tētahi whakapainga nui i te mātauranga.
Jaehyeon Kim et al.
MIT
Fast
en, zh, ja, ko
1GB
Kāore
Waihoki
MeloTTS
Free
Ko MeloTTS e MyShell.ai he puna TTS reo maha e tautoko ana i te reo Ingarihi (American, British, Indian, Australian), Spanish, French, Chinese, Japanese, me te Korean. He tere rawa, e mahi ana i te kupu i te tere o te wā tūturu i runga i te CPU anake. Kua hangaia a MeloTTS mō te whakamahinga whakanao, ā, e tautoko ana i te CPU me te GPU.
MyShell.ai
MIT
Fast
en, es, fr, zh, ja, ko
0.5GB (GPU optional)
Kāore
Waihoki
Bark
Standard
Ko Bark e Suno he tauira kupu-ki-rongoā i runga anō i te whakarerekētanga ka taea te whakaputa i te kōrero tino pono, i ngā reo maha, i ētahi atu oro pūoro pēnei i te pūoro, i te pōhēhētanga o te papamuri, i ngā pānga oro. Ka taea e ia te whakaputa i ngā whakawhitinga ā-waha pēnei i te māharahara, i te tūkinotanga, i te tūkinotanga. He nui ake i te 100 ngā whakaritenga kaikōrero me ngā reo 13+ e tautoko ana e Bark.
Suno
MIT
Slow
en, zh, fr, de, hi, it, ja, ko, pl, pt, ru, es, tr
5GB
Kāore
2x
Bark Small
Standard
He putanga iti ake o te tauira Bark ko Bark e whakawhiti ana i ētahi o ngā āhuatanga oro mō ngā tere whakahau tere ake me ngā hiahia pūmahara iti iho, e pupuri ana i te kaha o Bark ki te whakanao i te kōrero me ngā āhuatanga, te māharahara, me ngā reo maha.
Suno
MIT
Medium
en, zh, fr, de, hi, it, ja, ko, pl, pt, ru, es, tr
2GB
Kāore
2x
CosyVoice 2
Standard
Ko te CosyVoice 2 a Alibaba's Tongyi Lab e whiwhi ana i te āhua o te kōrero e ōrite ana ki te tangata me te pōturi iti rawa, e pai ana mō ngā taupānga wā tūturu. Ka whakamahia e ia tētahi huarahi whakarea tūturu mō te tāruatanga reo, ā, ka tautokona e ia te tāruatanga reo kore, te tāruatanga reo whakawhiti, me te whakahaere āhua o te āhua o te āhua o te āhua.
Alibaba (Tongyi Lab)
Apache 2.0
Medium
en, zh, ja, ko, fr, de, it, es
4GB
He
2x
Dia TTS
Standard
Ko Dia e Nari Labs he tauira kupu-ki-rongo 1.6B i hangaia mō te whakaputa i ngā kōrero maha. Ka taea e ia te whakaputa i ngā kōrero māori i waenganui i ngā kaikōrero e rua me te whakarerekētanga tika, me te kīanga ā-āhuatanga. He tino pai te Dia mō te waihanga i ngā ihirangi āhua podcast, ngā kōrero reo reo, me te AI whakawhitiwhitinga.
Nari Labs
Apache 2.0
Medium
en
4GB
Kāore
2x
Parler TTS
Standard
Ko te Parler TTS he tauira kupu-ki-rongo e whakamahi ana i ngā whakaahuatanga reo māori hei whakahaere i te kōrero i hangaia. Ehara i te kōwhiringa mai i ngā reo i whakaritea, ka whakaahuatia e koe te reo e hiahiatia ana e koe (hei tauira, "he reo wahine wera me tētahi āhuatanga British iti, e kōrero ana i te pōturi, i te mārama hoki") ā, ka whakaputaina e te Parler he kōrero e ōrite ana ki taua whakaahuatanga. Mā tēnei e āhei ai ki ngā taupānga auau.
Hugging Face
Apache 2.0
Medium
en
4GB
Kāore
2x
GLM-TTS
Standard
Ko te GLM-TTS na Zhipu AI he pūnaha kupu-ki-rongo i hangaia ki runga i te hanganga Llama me te ōritetanga rerenga. E whiwhi ana i te mokatere hapa ira iti rawa i waenganui i ngā tauira TTS pūtake tūwhera, ko te tikanga ka whakaputaina e ia te kōrero tino tika. E tautoko ana a GLM-TTS i te reo Ingarihi me te reo Hainamana me te tārua reo mai i ngā tauira oro 3-10 waeine.
Zhipu AI
GLM-4 License
Medium
en, zh
4GB
He
2x
IndexTTS-2
Standard
Ko te IndexTTS-2 he pūnaha tuhituhi-ki-te-kōrero hōhonu e tino pai ana ki te whakakotahi reo-kore me te whakahaere āhuahira-kore. Ka taea e ia te whakaputa kōrero me ngā āhuahira ā-āhuahira pēnei i te māharahara, i te pōhara, i te pōhara, i te pōhara rānei me te kore e hiahiatia he raraunga whakaakoranga ā-āhuahira. Ka whakamahia e te tauira ngā ira ā-āhuahira hei whakahaere tika i te kīanga ā-āhuahira o te kōrero i hangaia.
Index Team
Bilibili Model License
Medium
en, zh
4GB
He
2x
Spark TTS
Standard
Ko te Spark TTS na SparkAudio he tauira kupu-ki-whakaahua e hono ana i te tārua reo me te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua o te āhua.
SparkAudio
CC BY-NC-SA 4.0
Medium
en, zh
4GB
He
2x
GPT-SoVITS
Standard
Ko te GPT-SoVITS e whakakotahi ana i te tauira reo āhua GPT me te SoVITS (Singing Voice Inference mā te whakawhitinga me te whakakotahitanga) mō te tārua reo kaha-kore. Me te iti iho i te 5 sekone o te oro tohutoro, ka taea e ia te tārua tika i tētahi reo me te whakaputa reo hou i te wā e pupuri ana i ngā āhuatanga ahurei o te kaikōrero. He tino pai ki te kōrero me te whakakotahi reo.
RVC-Boss
MIT
Slow
en, zh, ja, ko
6GB
He
2x
Orpheus
Standard
Ko Orpheus he tauira kupu-ki-whakaahua nui e whiwhi ana i te kīanga ā-āhuatanga o te tangata. I whakaakona i runga i ngā raraunga kōrero maha ake i te 100,000 wā, e tino pai ana ki te whakaputa kōrero me ngā āhuatanga māori, te whakahua, me ngā kāhua kōrero. Ka taea e Orpheus te whakaputa kōrero e kore e taea te wehe i ngā pūkete tangata.
Canopy Labs
Llama 3.2 Community
Medium
en
4GB
Kāore
2x
Chatterbox
Premium
Ko te Chatterbox na Resemble AI he tauira tāruatanga oro-kore. Ka taea e ia te tārua i tētahi reo mai i tētahi tauira oro kotahi me te tika tino mōhio, kāore i te tango anake i te timbre engari ko te kāhua kōrero me ngā āhuatanga ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā.
Resemble AI
MIT
Medium
en
4GB
He
4x
Tortoise TTS
Premium
Ko te Tortoise TTS he pūnaha tuhituhi-ki-te-reo-maha e whakawhāiti ana i te āhua o te reo i runga i te tere. Ka whakamahia e ia te hanganga i whakaawetia e DALL-E hei waihanga i tētahi kōrero tino māori me te ōritetanga pai o te kōrero me te kaikōrero. Ahakoa he pōturi ake i ngā whirinoa maha, ka whakaputaina e te Tortoise ētahi o ngā kōrero tino mārama e wātea ana i roto i te pūnaha pūtake tūwhera.
James Betker
Apache 2.0
Slow
en
8GB
He
4x
StyleTTS 2
Premium
StyleTTS 2 e whiwhi ana i te hanganga TTS taumata- tangata mā te paheko i te whakawhānuitanga o te kāhua me te whakaakoranga ātete mā te whakamahi i ngā tauira reo kōrero nui. Ka whakaputaina e ia te kōrero tino māori i waenganui i ngā tauira kaikōrero kotahi, e whakataetae ana i ngā pūkete tangata. StyleTTS 2 e whakamahi ana i te tauira kāhua i runga anō i te whakawhānuitanga hei tango i te awhe katoa o te rerekētanga o te reo tangata.
Columbia University
MIT
Medium
en
4GB
Kāore
4x
OpenVoice
Premium
E āhei ana a OpenVoice e MyShell.ai ki te tārua reo tere me te whakahaere matatini i runga i te kāhua reo, i te āhua, i te āhua, i te āhua, i te wā, i te āhua. Ka taea e ia te tārua i tētahi reo mai i tētahi rīpene orooro poto me te whakaputa kōrero i ngā reo maha i te pupuri i te tuakiri o te kaikōrero. Ka mahi hoki a OpenVoice hei kaiwhakarere reo, e whakaae ana ki te huringa reo i te wā tūturu.
MyShell.ai / MIT
MIT
Medium
en, zh, ja, ko, fr, de, es, it
4GB
He
4x
Qwen3 TTS
Standard
Ko Qwen3-TTS he tauira tuhi-ki-te-kōrero tauine 1.7 miriona mai i te rōpū Qwen o Alibaba. E toru ngā āhuatanga e tautoko ana i a ia: ngā reo i whakaritea i mua me te mana ā-āhuatanga (9 ngā kaikōrero), te tārua reo mai i ngā waeine 3 anake o te oro, me tētahi āhuatanga hoahoa reo motuhake e whakaahua ana i te reo e hiahiatia ana e koe i roto i te reo māori.
Alibaba (Qwen)
Apache 2.0
Medium
en, zh, ja, ko, de, fr, ru, pt, es, it
7GB
He
2x
Sesame CSM
Premium
Ko te Sesame CSM (Model Speech Conversational) he tauira taurearea kotahi mano, kua hangaia hei whakaputa kōrero ā-waha. Ka tauiratia e ia ngā tauira tūturu o te kōrero tangata tae atu ki te wā whakarerekētanga, ngā urupare ā-roto, ngā urupare ā-āhuatanga, me te rerenga kōrero.
Sesame
Apache 2.0
Slow
en
8GB
Kāore
4x
Chatterbox Turbo
Standard
Ko te Chatterbox Turbo na Resemble AI he whakawhānuitanga tohuāhua 350M ki te Chatterbox, e tuku ana ki te tere o te wā tūturu o te 6x me te ātetetanga o te 200ms. E tautoko ana i ngā tohu paralinguistic pēnei i te [laugh], [cough], me te [chuckle] i roto i te kupu. Kei roto ko te Perth watermarking i ngā oro katoa i hangaia mō te whai i te take.
Resemble AI
MIT
Fast
en
2GB
He
2x
Zonos
Standard
Ko te Zonos v0.1 na Zyphra he tauira tohuāhua 1.6B e whakaatu ana i te mana ā-āhuatanga mārō me ngā kāwai mō te aroha, te pōharatanga, te pōharatanga, te pōharatanga, me te whakamātautau. E whakarato ana i tētahi Transformer me tētahi tāupe SSM hōu (tauira mokowā ā-kāwanatanga). I whakaakona i runga i ngā wā 200K+ o te kōrero maha me te tārua reo-kore mai i ngā waeine 10-30 o te oro tohutoro.
Zyphra
Apache 2.0
Medium
en, ja, zh, fr, de
6GB
He
2x
Dia 2
Standard
Ko te Dia2 a Nari Labs he whakawhānuitanga-tūturu ki te Dia, e wātea ana i roto i ngā tāupe tohu 1B me te 2B. Ka tīmata ki te whakawhanake i te oro mai i ngā tohu tuatahi, e pai ai mō ngā māngai reo wā-tūturu me ngā pūwhitinga kōrero-ki-te-kōrero. E tautoko ana i te kōrerorero maha me ngā tohu [S1] / [S2] me ngā tohu paralinguistic pēnei i te (laughs), (coughs).
Nari Labs
Apache 2.0
Fast
en
4GB
Kāore
2x
VoxCPM
Standard
Ko te VoxCPM 1.5 na OpenBMB he tauira TTS kore tohu hou e mahi ana i roto i te mokowā tūturu ehara i te tohu motuhake. Ka whakaputaina e ia i te oro 44.1kHz, e tautoko ana i te tārua reo kore-kōrero mai i te 3-10 sekone, ā, ka pupuri i te ōritetanga puta noa i ngā wāhanga. Ka taea e te tārua reo te hoatu i tētahi reo Ingarihi ki te kōrero Hainamana, ā, ko te āhua anō.
OpenBMB
Apache 2.0
Fast
en, zh
4GB
He
2x
OuteTTS
Free
E whakaroa ana e te OuteTTS ngā tauira reo nui me ngā āheinga kupu-ki-whakaahua i te wā e pupuri ana i te hanganga taketake. E tautoko ana i ngā taupoki maha tae atu ki a llama.cpp (CPU/GPU), Hugging Face Transformers, ExLlamaV2, VLLM, ā, ko te whakawāteatanga whakangākau mā Transformers.js. He āhuahira te tārua reo kore-pōti mā ngā tātai kōrero i tiakina hei JSON.
OuteAI
Apache 2.0
Fast
en
2GB
He
Waihoki
TADA
Standard
TADA (Text-Acoustic Dual Alignment) e Hume AI he tauira TTS whakahauhau e whakakore ana i ngā hallucinations mā tētahi hoahoa tapawhā hōu i hangaia ki Llama 3.2. Kei te wātea i roto i ngā tāupe 1B (English) me te 3B (maha-reo), ka tae mai a TADA ki tētahi RTF o te 0.09 — 5x tere ake i ngā tauira TTS i runga i te LLM. E tautoko ana i te 700 waeine o te horopaki oro, ā, ka whakaputaina he kōrero ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā-ā
Hume AI
MIT
Fast
en
5GB
Kāore
2x
VibeVoice
Standard
E rua nga momo VibeVoice a Microsoft: he tauira 1.5B mō ngā ihirangi ā-rohe roa (tata ki ngā minu 90, 4 ngā kaikōrero) me tētahi tauira 0.5B o te wā tūturu mō te whakawhitinga me te ātete reo tuatahi ~200ms. Ko te momo 1.5B e tino pai ana i ngā podcast me ngā pukapuka oro me te ōritetanga o te kaikōrero i ngā whakawhitinga roa. Whakama: I tangohia e Microsoft te waehere TTS mai i te puna, ā, ko te oro i whakaputaina e whakauru ana i ngā whakawāteatanga AI.
Microsoft
MIT
Fast
en, zh
4GB
Kāore
2x
Pocket TTS
Free
Ko te Pocket TTS a Kyutai (ngā kaihanga o Moshi) he tauira kupu-ki-whakaahua 100M tauine-ki-whakaahua e whakarewa ana i runga ake i tōna taumahatanga. Ka mahi tika i runga i te CPU, e tautoko ana i te tārua reo kore-kōrero mai i tētahi tauira oro kotahi, ā, ka whakaputaina he kōrero māori. Ko te rahi o te tauira iti e tino pai ana mō te whakawhānui i te pito me ngā taiao rawa iti.
Kyutai
MIT
Fast
en, fr
1GB
He
Waihoki
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
Kāore
Waihoki
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
He
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
He
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
He
4x
Kokoro
Waihoki
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
Waihoki
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
Waihoki
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
Waihoki
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
Waihoki
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
Waihoki
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
Waihoki
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
Paerewa
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
Kāore
Bark Small
Paerewa
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
Kāore
CosyVoice 2
Paerewa
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
He
Dia TTS
Paerewa
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
Kāore
Parler TTS
Paerewa
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
Kāore
GLM-TTS
Paerewa
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
He
IndexTTS-2
Paerewa
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
He
Spark TTS
Paerewa
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
He
GPT-SoVITS
Paerewa
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
He
Orpheus
Paerewa
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
Kāore
Qwen3 TTS
Paerewa
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
He
Chatterbox Turbo
Paerewa
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
He
Zonos
Paerewa
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
He
Dia 2
Paerewa
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
Kāore
VoxCPM
Paerewa
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
He
TADA
Paerewa
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
Kāore
VibeVoice
Paerewa
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
Kāore
CosyVoice3
Paerewa
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
He
Chatterbox
Whakawhiwhinga
Chatterbox by Resemble AI is a cutting-edge zero-shot voice cloning model. It can replicate any voice from a single audio sample with remarkable accuracy, capturing not just the timbre but also the speaking style and emotional nuances. Chatterbox also features fine-grained emotion control, allowing you to adjust the emotional tone of the generated speech independently from the voice identity.
Resemble AI
MIT
Medium
en
He
4GB
4x
Tortoise TTS
Whakawhiwhinga
Tortoise TTS is an autoregressive multi-voice text-to-speech system that prioritizes audio quality over speed. It uses DALL-E-inspired architecture to generate highly natural speech with excellent prosody and speaker similarity. While slower than many alternatives, Tortoise produces some of the most realistic synthetic speech available in the open-source ecosystem.
James Betker
Apache 2.0
Slow
en
He
8GB
4x
StyleTTS 2
Whakawhiwhinga
StyleTTS 2 achieves human-level TTS synthesis by combining style diffusion with adversarial training using large speech language models. It generates the most natural sounding speech among single-speaker models, rivaling human recordings. StyleTTS 2 uses diffusion-based style modeling to capture the full range of human speech variation.
Columbia University
MIT
Medium
en
Kāore
4GB
4x
OpenVoice
Whakawhiwhinga
OpenVoice by MyShell.ai enables instant voice cloning with granular control over voice style, emotion, accent, rhythm, pauses, and intonation. It can clone a voice from a short audio clip and generate speech in multiple languages while maintaining the speaker identity. OpenVoice also functions as a voice converter, allowing real-time voice transformation.
MyShell.ai / MIT
MIT
Medium
en, zh, ja, ko, fr, de, es, it
He
4GB
4x
Sesame CSM
Whakawhiwhinga
Sesame CSM (Conversational Speech Model) is a 1 billion parameter model designed specifically for generating conversational speech. It models the natural patterns of human conversation including turn-taking timing, backchannel responses, emotional reactions, and conversational flow. CSM generates audio that sounds like a natural human conversation rather than synthetic speech.
Sesame
Apache 2.0
Slow
en
Kāore
8GB
4x
MOSS-TTS
Whakawhiwhinga
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
He
16GB
4x
MegaTTS3
Whakawhiwhinga
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
He
8GB
4x
Te ripanga whakataurite tauira
| Kāhua | kaiwhakawhanake: | Te āhua | Kāwai: | Āhuatanga | reo | Ko te tāruatanga reo | VRAM | Whakawhiwhinga: | pūtea | |
|---|---|---|---|---|---|---|---|---|---|---|
| Kokoro | Hexgrad | Free | Fast | 11 | 1.5GB | Apache 2.0 | Waihoki | Ka whakamahia | ||
| Piper | Rhasspy | Free | Fast | 31 | 0 (CPU only) | MIT | Waihoki | Ka whakamahia | ||
| VITS | Jaehyeon Kim et al. | Free | Fast | 4 | 1GB | MIT | Waihoki | Ka whakamahia | ||
| MeloTTS | MyShell.ai | Free | Fast | 6 | 0.5GB (GPU optional) | MIT | Waihoki | Ka whakamahia | ||
| Bark | Suno | Standard | Slow | 13 | 5GB | MIT | 2 | Ka whakamahia | ||
| Bark Small | Suno | Standard | Medium | 13 | 2GB | MIT | 2 | Ka whakamahia | ||
| CosyVoice 2 | Alibaba (Tongyi Lab) | Standard | Medium | 8 | 4GB | Apache 2.0 | 2 | Ka whakamahia | ||
| Dia TTS | Nari Labs | Standard | Medium | 1 | 4GB | Apache 2.0 | 2 | Ka whakamahia | ||
| Parler TTS | Hugging Face | Standard | Medium | 1 | 4GB | Apache 2.0 | 2 | Ka whakamahia | ||
| GLM-TTS | Zhipu AI | Standard | Medium | 2 | 4GB | GLM-4 License | 2 | Ka whakamahia | ||
| IndexTTS-2 | Index Team | Standard | Medium | 2 | 4GB | Bilibili Model License | 2 | Ka whakamahia | ||
| Spark TTS | SparkAudio | Standard | Medium | 2 | 4GB | CC BY-NC-SA 4.0 | 2 | Ka whakamahia | ||
| GPT-SoVITS | RVC-Boss | Standard | Slow | 4 | 6GB | MIT | 2 | Ka whakamahia | ||
| Orpheus | Canopy Labs | Standard | Medium | 1 | 4GB | Llama 3.2 Community | 2 | Ka whakamahia | ||
| Chatterbox | Resemble AI | Premium | Medium | 1 | 4GB | MIT | 4 | Ka whakamahia | ||
| Tortoise TTS | James Betker | Premium | Slow | 1 | 8GB | Apache 2.0 | 4 | Ka whakamahia | ||
| StyleTTS 2 | Columbia University | Premium | Medium | 1 | 4GB | MIT | 4 | Ka whakamahia | ||
| OpenVoice | MyShell.ai / MIT | Premium | Medium | 8 | 4GB | MIT | 4 | Ka whakamahia | ||
| Qwen3 TTS | Alibaba (Qwen) | Standard | Medium | 10 | 7GB | Apache 2.0 | 2 | Ka whakamahia | ||
| Sesame CSM | Sesame | Premium | Slow | 1 | 8GB | Apache 2.0 | 4 | Ka whakamahia | ||
| Chatterbox Turbo | Resemble AI | Standard | Fast | 1 | 2GB | MIT | 2 | Ka whakamahia | ||
| Zonos | Zyphra | Standard | Medium | 5 | 6GB | Apache 2.0 | 2 | Ka whakamahia | ||
| Dia 2 | Nari Labs | Standard | Fast | 1 | 4GB | Apache 2.0 | 2 | Ka whakamahia | ||
| VoxCPM | OpenBMB | Standard | Fast | 2 | 4GB | Apache 2.0 | 2 | Ka whakamahia | ||
| OuteTTS | OuteAI | Free | Fast | 1 | 2GB | Apache 2.0 | Waihoki | Ka whakamahia | ||
| TADA | Hume AI | Standard | Fast | 1 | 5GB | MIT | 2 | Ka whakamahia | ||
| VibeVoice | Microsoft | Standard | Fast | 2 | 4GB | MIT | 2 | Ka whakamahia | ||
| Pocket TTS | Kyutai | Free | Fast | 2 | 1GB | MIT | Waihoki | Ka whakamahia | ||
| Kitten TTS | KittenML | Free | Fast | 1 | 0GB | Apache 2.0 | Waihoki | Ka whakamahia | ||
| CosyVoice3 | Alibaba (FunAudioLLM) | Standard | Fast | 9 | 4GB | Apache 2.0 | 2 | Ka whakamahia | ||
| MOSS-TTS | OpenMOSS | Premium | Medium | 19 | 16GB | Apache 2.0 | 4 | Ka whakamahia | ||
| MegaTTS3 | ByteDance | Premium | Slow | 2 | 8GB | Apache 2.0 | 4 | Ka whakamahia |
Ko te pūwāhi kupu AI tino whānui ki te kōrerorero
He aha te kōwhiringa a TTS.ai mō te kupu ki te kōrero?
TTS.ai e whakakotahi ana te ao
Ko ia tauira he pūtake tūwhera i raro i te MIT, Apache 2.0, he whakaaetanga ōrite rānei, e whakaū ana i ōna mana hokohoko katoa hei whakamahi i te oro i hangaia i roto i ōna kaupapa. Mēnā e hiahiatia ana e koe he whakakotahitanga tere, māmā rānei mō ngā taupānga wā tūturu, te huaputa mātauranga rānei mō ngā pukapuka oro me ngā podcast, he tauira tika a TTS.ai mō ia take whakamahi.
Kāhua wātea, kāore he tatau e hiahiatia ana
Ka tīmata i te wā kotahi ki ngā tauira TTS wātea e toru: Piper (āhua tere, māmā), VITS (whakahaeretanga ā-ira nui), me MeloTTS (whakahaeretanga reo maha). Kāore he whakaingoatanga, kāore he kāri pūtea, kāore he tepe i runga i ngā whakatupuranga. Ko ngā tauira wātea e tautoko ana i te reo Ingarihi me ētahi atu reo maha me ngā huaputa pūoro māori e tika ana mō te nuinga o ngā taupānga.
Ka whakateretia te tukanga GPU
Ko ngā tauira TTS katoa e haere ana i runga i ngā GPU NVIDIA motuhake mō ngā wā whakawhanake tere, ōrite. Ko ngā tauira wātea e whakaputa reo ana i raro iho i te 2 sekone. Ko ngā tauira paerewa pēnei i a Kokoro, CosyVoice 2, me Bark te nuinga o te 3-5 sekone. Ko ngā tauira utu me te āhuatanga tiketike rawa, pēnei i a Tortoise me Chatterbox, e mahi ana i roto i te 5-15 sekone, i runga anō i te roanga o te kupu.
30+ reo kua tautokona
Ka whakaputa kōrero i ngā reo neke atu i te 30 tae atu ki te reo Ingarihi, Paniora, Wīwī, Tiamana, Itari, Portuguese, Hainamana, Hapanihi, Koreana, Arabic, Hindi, Rūhia, me ētahi atu. He maha ngā tauira e tautoko ana i te whakawhiti-reo, ko te tikanga ka taea e koe te whakaputa kōrero i roto i tētahi reo kāore anō kia whakaakona te reo taketake. Ko CosyVoice 2 me GPT-SoVITS e tino pai ana i te tārua reo whakawhiti-reo.
Ka whakaritea e te kaiwhakawhanake
Ka whakaurua a TTS.ai ki ōna taupānga me a tātau OpenAI-hoatu REST API. He wāhi mutunga kotahi mō ngā tauira 20+ katoa. Python, JavaScript, cURL, me Go SDKs. Whakawhiwhinga tautoko mō ngā taupānga wā tūturu. Whakaputanga rōpū mō te whakawhanaketanga ihirangi nui. Webhooks mō ngā mōhiohio async. E wātea ana ki ngā mahere Pro me Enterprise.
E pā ana ngā pātai
What could we improve? Your feedback helps us fix issues.
Ka tīmata te tahuritanga o te kupu ki te kōrero ināianei
Ka hono ki ngā mano o ngā kaihanga e whakamahi ana i te TTS.ai. Ka whiwhi 15,000 ngā pūāhua wātea me tētahi pūtake hou. Ka wātea ngā tauira wātea me te kore whakaingoatanga.