Tọghata ngwe ka ọsụsọ Ndịna AI Ụdị Nsụgharị Klọnsị ụda Nhazi ụda Marketplace Oge
Ndesịta ihenhọrọ ndị ahụ

TTS Arena — AI Voice Model Leaderboard

Tụkwasị 20+ ngwe-na-asụsụ model. Official benchmarks, community ratings, na side-by-side comparison.

Akaụntụ
Anyị enweghị ụda TTS n'asụsụ gị kemgbe. Meekwa ka anyị tinye gị! Kpọnye ụda gị

Ndesịta ihenhọrọ ndị ahụ

Tinye ngwe, họrọ móòdù abụọ, nakwa belata nsonaazụ. Free-tier móòdù chọrọ akaụntụ ọbụla.

Free models na-arụ ọrụ na-enweghị akaụntụ. Akaụntụ iji tụnyere ụdị premium.

Móòdù Leaderboard

# Móòdù Òfisíèlì Òtù Ọnụọgụgụ gị Nhazi Tier
1
Kokoro
Kokoro
Lightweight 82M parameter model delivering studio-quality speech with blazing-fast inference.
82M 1200h 2024
4.8 /5 5.0 /5
1 Ọnụọgụgụ
fast Free
2
CosyVoice 2
CosyVoice 2
Alibaba's scalable streaming TTS with human-parity naturalness and near-zero latency.
300M 200000h 2024
4.26 /5 Enweghị ụtụ isi
medium Standard
3
Chatterbox
Chatterbox
State-of-the-art zero-shot voice cloning with emotion control from Resemble AI.
300M 2025
4.25 /5 Enweghị ụtụ isi
medium Premium
4
StyleTTS 2
StyleTTS 2
Human-level text-to-speech through style diffusion and adversarial training.
100M 585h 2024
4.23 /5 Enweghị ụtụ isi
medium Premium
5
Piper
Piper
A fast, local neural text to speech system optimized for Raspberry Pi and embedded devices.
15M 2023
4.15 /5 Enweghị ụtụ isi
fast Free
6
MeloTTS
MeloTTS
High-quality multilingual text-to-speech that runs on CPU with minimal latency.
25M 2024
4.13 /5 Enweghị ụtụ isi
fast Free
7
Dia TTS
Dia TTS
Multi-speaker dialog generation model that creates natural conversations between speakers.
1.6B 2024
4.09 /5 Enweghị ụtụ isi
medium Standard
8
VITS
VITS
Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech.
25M 585h 2021
4.0 /5 Enweghị ụtụ isi
fast Free
9
Orpheus
Orpheus
Human-level emotional TTS model trained on 100K hours of speech data.
3B 100000h 2025
4.0 /5 Enweghị ụtụ isi
medium Standard
10
OpenVoice
OpenVoice
Instant voice cloning with granular control over style, emotion, and accent.
300M 2024
4.0 /5 Enweghị ụtụ isi
medium Premium
11
IndexTTS-2
IndexTTS-2
Zero-shot TTS with fine-grained emotion control and high expressiveness.
300M 2025
3.91 /5 Enweghị ụtụ isi
medium Standard
12
Spark TTS
Spark TTS
Voice cloning TTS with controllable emotion and speaking style via prompts.
500M 2025
3.9 /5 Enweghị ụtụ isi
medium Standard
13
Parler TTS
Parler TTS
Describe the voice you want in natural language and Parler generates matching speech.
880M 45000h 2024
3.83 /5 Enweghị ụtụ isi
medium Standard
14
Tortoise TTS
Tortoise TTS
Multi-voice text-to-speech focused on quality with autoregressive architecture.
400M 50000h 2022
3.7 /5 Enweghị ụtụ isi
slow Premium
15
Bark
Bark
Transformer-based text-to-audio model that generates realistic speech, music, and sound effects.
350M 100000h 2023
3.57 /5 Enweghị ụtụ isi
slow Standard
16
Bark Small
Bark Small
Lighter version of Bark with faster inference and lower memory usage.
150M 100000h 2023
— Enweghị ụtụ isi
medium Standard
17
Indic Parler TTS
Indic Parler TTS
High-quality speech for 8+ Indian languages with natural-language voice control.
900M 8000h 2024
— Enweghị ụtụ isi
slow Standard
18
KhanomTan TTS
KhanomTan TTS
Thai-first text-to-speech with a choice of speaker voices.
85M 100h 2023
— Enweghị ụtụ isi
fast Standard
19
GPT-SoVITS
GPT-SoVITS
Few-shot voice cloning TTS that replicates any voice from just 5 seconds of audio.
200M 2024
— Enweghị ụtụ isi
slow Standard
20
Qwen3 TTS
Qwen3 TTS
Alibaba's multilingual TTS with preset voices and voice design from text.
1.7B 2025
— Enweghị ụtụ isi
medium Standard
21
VieNeu-TTS-v2
VieNeu-TTS-v2
Vietnamese + English code-switching TTS with 7 preset voices and zero-shot voice cloning. CPU-only, no GPU required.
0.3B 10000h 2026
— Enweghị ụtụ isi
fast Standard
22
Sesame CSM
Sesame CSM
Conversational speech model generating natural dialogue with appropriate timing and emotion.
1B 2025
— Enweghị ụtụ isi
slow Premium
23
Chatterbox Turbo
Chatterbox Turbo
Faster Chatterbox with sub-200ms latency and paralinguistic tags for laughs, coughs, and more.
350M 2025
— Enweghị ụtụ isi
fast Standard
24
VoxCPM
VoxCPM
Tokenizer-free TTS producing 44.1kHz audio with context-aware paragraph consistency.
500M 1800000h 2025
— Enweghị ụtụ isi
fast Standard
25
Kani TTS 2
Kani TTS 2
Ultra-lightweight 400M English TTS model running in just 3GB VRAM.
400M 10000h 2026
— Enweghị ụtụ isi
fast Free
26
OuteTTS
OuteTTS
LLM-based TTS that runs on CPU, GPU, or browser via llama.cpp and Transformers.js.
1B 5000h 2025
— Enweghị ụtụ isi
slow Free
27
VibeVoice
VibeVoice
Microsoft's multi-speaker long-form TTS generating up to 90 minutes with 4 distinct speakers.
1.5B 100000h 2025
— Enweghị ụtụ isi
fast Standard
28
Pocket TTS
Pocket TTS
Lightweight 100M parameter model by Kyutai with voice cloning from a single sample.
100M 50000h 2025
— Enweghị ụtụ isi
fast Free
29
Kitten TTS
Kitten TTS
Ultra-lightweight TTS under 80MB. Runs on CPU without GPU.
80M 2025
— Enweghị ụtụ isi
fast Free
30
CosyVoice3
CosyVoice3
Next-generation multilingual TTS with bi-streaming, emotion control, and zero-shot voice cloning.
500M 200000h 2025
— Enweghị ụtụ isi
fast Standard
31
NAMAA Saudi TTS
NAMAA Saudi TTS
First open Saudi-Arabic TTS. Native Saudi dialect with Chatterbox-quality voice cloning.
300M 2026
— Enweghị ụtụ isi
medium Standard
32
Darwin TTS
Darwin TTS
Cross-modal Qwen3-TTS variant with FFN weights blended from the Qwen3-1.7B language model for sharper multilingual cloning.
2.1B 2026
— Enweghị ụtụ isi
medium Standard
33
MOSS-TTSD
MOSS-TTSD
Multi-speaker dialogue continuation model — generate podcast-style conversations with up to 5 speakers and 60 minutes of coherent audio.
7B 2026
— Enweghị ụtụ isi
medium Standard
34
Ming-Omni TTS
Ming-Omni TTS
Compact 0.5B omni-modal speech model from inclusionAI with high-fidelity 44.1kHz output and zero-shot voice cloning.
500M 2026
— Enweghị ụtụ isi
medium Free
35
MOSS-TTS Nano
MOSS-TTS Nano
Tiny 100M MOSS-TTS variant — same architecture, 80x smaller, free-tier latency.
100M 500000h 2026
— Enweghị ụtụ isi
fast Free
36
FreyaTTS
FreyaTTS
Turkish-only speech at 48 kHz from a compact flow-matching model.
183M 2026
— Enweghị ụtụ isi
fast Free

Ndesịta ihenhọrọ nke benchmark

Official TTS.ai benchmark scores across three dimensions: naturalness, accuracy, and speed.

KokoroKokoro

Free
Ọdịnaya 4.8/5
Nhazi 4.7/5
Nhazi 4.9/5
Nhazi niile 4.8/5

CosyVoice 2CosyVoice 2

Standard
Ọdịnaya 4.5/5
Nhazi 4.4/5
Nhazi 3.8/5
Nhazi niile 4.26/5

ChatterboxChatterbox

Premium
Ọdịnaya 4.7/5
Nhazi 4.5/5
Nhazi 3.4/5
Nhazi niile 4.25/5

StyleTTS 2StyleTTS 2

Premium
Ọdịnaya 4.5/5
Nhazi 4.3/5
Nhazi 3.8/5
Nhazi niile 4.23/5

PiperPiper

Free
Ọdịnaya 3.5/5
Nhazi 4.2/5
Nhazi 4.95/5
Nhazi niile 4.15/5

MeloTTSMeloTTS

Free
Ọdịnaya 3.8/5
Nhazi 4.1/5
Nhazi 4.6/5
Nhazi niile 4.13/5

Dia TTSDia TTS

Standard
Ọdịnaya 4.6/5
Nhazi 4.3/5
Nhazi 3.2/5
Nhazi niile 4.09/5

VITSVITS

Free
Ọdịnaya 3.4/5
Nhazi 4.0/5
Nhazi 4.8/5
Nhazi niile 4.0/5

OrpheusOrpheus

Standard
Ọdịnaya 4.3/5
Nhazi 4.1/5
Nhazi 3.5/5
Nhazi niile 4.0/5

OpenVoiceOpenVoice

Premium
Ọdịnaya 4.0/5
Nhazi 4.1/5
Nhazi 3.9/5
Nhazi niile 4.0/5

IndexTTS-2IndexTTS-2

Standard
Ọdịnaya 4.3/5
Nhazi 4.1/5
Nhazi 3.2/5
Nhazi niile 3.91/5

Spark TTSSpark TTS

Standard
Ọdịnaya 4.2/5
Nhazi 4.0/5
Nhazi 3.4/5
Nhazi niile 3.9/5

Parler TTSParler TTS

Standard
Ọdịnaya 4.1/5
Nhazi 3.9/5
Nhazi 3.4/5
Nhazi niile 3.83/5

Tortoise TTSTortoise TTS

Premium
Ọdịnaya 4.6/5
Nhazi 4.4/5
Nhazi 1.8/5
Nhazi niile 3.7/5

BarkBark

Standard
Ọdịnaya 4.2/5
Nhazi 3.8/5
Nhazi 2.5/5
Nhazi niile 3.57/5

Nhazi

Nhazi ule

  • Háàrị̀: 4x NVIDIA Tesla P40 (24GB VRAM otu), 96GB zuru ezu
  • Nkọwa: 5 standardized passages covering different speech patterns (nkọwa, okwu, teknụzụ, mmetụta uche, asụsụ dị iche iche)
  • Nhazi: Automatic metrics (MOS estimation, WER, RTF) jikọtara ya na human listening tests
  • Ọrụ: Móòdù ọbụla enyochaala ugboro 10 n'ụbọchị ọbụla, ọnụọgụgụ n'ụbọchị

Nhazi

  • Nhazi (40%): Prosody, intonation, rhythm, emotions - olee otú ọ ga-esi dị mmadụ ụtọ?
  • Nhazi (30%): Nkọwa ziri ezi, ọnụọgụgụ njehie okwu, nghọta
  • Ogo (30%): Real-time factor (audio seconds / generation seconds). Ọnụọgụgụ dị elu = ngwa ngwa.
  • Ogo: Wepụtara ọnụọgụgụ: 0.4 x Naturalness + 0.3 x Accuracy + 0.3 x Speed

Ntụziaka: Benchmarks na-egosipụta mmepe na haịdrọịd na ngwe nnwale anyị. Ọdịnaya nke ụwa n'ozuzu ya nwere ike ịdị iche iche site n'input ngwe, asụsụ, na nhọpụta ụda. Ọnụọgụgụ okpukperechi na-enye akara n'ozuzu ya site n'ịrụ ọrụ n'ozuzu ya.

Ajụjụ ndị a na-ajụkarị

The TTS Arena bụ a leaderboard nke rankings AI text-to-speech models based on official benchmark tests and community ratings. Compare models side-by-side, listen to samples, and vote for the ones that sound best to you.

Anyị na-arụ ọrụ nnyocha nkịtị n'ụdị ọ bụla na-eji ngwe ndị ahụ, kọmpụta, nakwa ihenhọrọ nlebara anya. Ọnụọgụgụ na-ekpuchi ọdịdị (otú ọ dị ka ọ dị ka mmadụ), n'ụzọ ziri ezi (ịsụgharị na nghọta), na ọsọ (oge mmepe). Nnyocha niile na-eji GPU sava anyị na NVIDIA Tesla P40 GPUs.

Ee! Pịa ndị kpakpando n'ebe ọbụla n'ime móòdù ọbụla ka ị ráàtụ̀ ya site na 1 ruo 5. I kwesịrị ịbanye n'ime ka ị na-ahọrọ. Ráàtụ̀ gị na-enyere aka n'ihe fọrọ nke nta ka ọ bụrụ obodo ahụ a na-egosi na leaderboard. I nwere ike ịgbanwe ráàtụ̀ gị mgbe ọbụla.

Tinye ngwe ọbụla, họrọ móòdù abụọ, wee pịa Tụkwasị. Módù abụọ ahụ na-eweta okwu site n'okwu ahụ n'otu oge. Gụọ ha abụọ ma ọ bụ họrọ nke dị mma. Ntụkwasị a na-enyere aka ịkọwa módù kacha mma maka mkpa gị.

Nnwere onwe na-enyocha otú ụda okwu dị ka mmadụ si dị (prosody, intonation, rhythm). Nnwere onwe na-enyocha n'ụzọ ziri ezi na nghọta. Nnwere ike na-enyocha otú ngwa ngwa móòdù na-ebipụta ụda dị ka ọbụla. Nnwere onwe bụ ọnụọgụgụ nke metrik niile.

Models na-enweghị benchmark scores bụ ma ọ bụ nke ọfụụ agbakwunyere ma na-atụ anya ịtụle, mọọbụ chọrọ nhazi pụrụ iche (dị ka gated access tokens) nke na-akwụsị. Ntụle obodo ka nọ n'ọrụ maka models ndị a.

Nhazi nke n'aka ndị ọrụ na-akpụgharị mgbe móòdù na-enweta mmegharị ndị dị mkpa mọọbụ mgbe mòdù ọhụrụ na-agbakwunyere. Nhazi nke obodo na-akpụgharị n'oge ọbụla dịka ndị ọrụ na-ahọrọ. Data leaderboard a na-echekwa maka minit 5 maka mmezi.

Free-tier models (Kokoro, Piper, VITS, MeloTTS) na-eburu ọbụna premium surcharge na-ewepụ gị free allowance. Standard models na-eji 2x characters (eg, 1,000 characters nke ngwe na-akwụ ụgwọ 2,000 characters site na ego gị). Premium models na-eji 4x characters na-enyekarị ihe dị elu ma ọ bụ ihe nzuzo dị ka ụda cloning.

Maka ihenhọrọ ndị a na-ejikarị, Kokoro (free tier) na-enyekwa mmanya dị mma. Maka ịkọsa ụda, jiri Chatterbox ma ọ bụ CosyVoice 2. Maka ihenhọrọ ndị na-asụ asụsụ dị iche iche, MeloTTS ma ọ bụ CosyVoice 2. Maka ntụgharị okwu dị iche iche, Bark ma ọ bụ Dia. Jiri ihenhọrọ ịgbanwee iji hụ na ngwe gị dị iche iche.

Ee, ị nwere ike ịmepụta nakwa ịtụle ụda site na ụdị ọbụla abụọ na-enweghị akaụntụ na-eji ụdị n'efu. Ikpe n'ihe ngosi na-achọ akaụntụ n'efu. Ntụle ụdị premium chọrọ akara.

Anyị na-arụsi ọrụ ike maka objectivity site na iji standardized test texts, identical hardware, na consistent evaluation criteria n'akụkụ niile models. Community ratings na-enye a ọzọ ọkaibe akara. Our methodology bụ kọwapụtara na Benchmark Methodology section n'okpuru.

Ụdị ndị ahụ a haziri ha site n'aka onyeisi benchmark zuru ezu, mgbe ahụ site n'aka ọnụọgụgụ obodo dị ka tiebreaker. Ụdị ndị na-enweghị benchmarks a haziri ha n'okpuru ndị nwere benchmarks, a haziri ha site n'aka ọnụọgụgụ obodo.
5.0/5 (1)

Gịnị ka anyị ga-eme ka ọ dịrị mma? Ntụziaka gị na-enyere anyị aka idozi nsogbu.

Chọ̀ọ́ ụda gị zuru ezu

Nwalee ụdị ọbụla n'efu na Kokoro, Piper, VITS, mọọbụ MeloTTS. Achọrọ akaụntụ ọbụla.

Akaụntụ Gosi ọnụahịa