TTS Arena — AI Voice Model Leaderboard (an angle)
Konpare 20 + modèl tèks-a-parole. Rekòmandasyon ofisyèl, ratings kominote, ak konparezon bò-a-bò.
Nou fè
Vann Voy ou
Konparezon
Tape tèks, chwazi de modèl, epi konpare rezilta yo. Modèl gratis pa bezwen okenn kont.
Modèl gratis travay san yon kont. Enskri pou konpare modèl prim.
Model Leaderboard
| # | Modèl | Ofisyèl | Kominote | Rating ou | Vitès | Nivo |
|---|---|---|---|---|---|---|
| 1 |
Kokoro
Lightweight 82M parameter model delivering studio-quality speech with blazing-fast inference.
82M
1200h
2024
|
4.8 /5 |
5.0
/5
1 vote
|
fast | Free | |
| 2 |
CosyVoice 2
Alibaba's scalable streaming TTS with human-parity naturalness and near-zero latency.
300M
200000h
2024
|
4.26 /5 | Pa gen vòt | medium | Standard | |
| 3 |
Chatterbox
State-of-the-art zero-shot voice cloning with emotion control from Resemble AI.
300M
2025
|
4.25 /5 | Pa gen vòt | medium | Premium | |
| 4 |
StyleTTS 2
Human-level text-to-speech through style diffusion and adversarial training.
100M
585h
2024
|
4.23 /5 | Pa gen vòt | medium | Premium | |
| 5 |
Piper
A fast, local neural text to speech system optimized for Raspberry Pi and embedded devices.
15M
2023
|
4.15 /5 | Pa gen vòt | fast | Free | |
| 6 |
MeloTTS
High-quality multilingual text-to-speech that runs on CPU with minimal latency.
25M
2024
|
4.13 /5 | Pa gen vòt | fast | Free | |
| 7 |
Dia TTS
Multi-speaker dialog generation model that creates natural conversations between speakers.
1.6B
2024
|
4.09 /5 | Pa gen vòt | medium | Standard | |
| 8 |
VITS
Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech.
25M
585h
2021
|
4.0 /5 | Pa gen vòt | fast | Free | |
| 9 |
Orpheus
Human-level emotional TTS model trained on 100K hours of speech data.
3B
100000h
2025
|
4.0 /5 | Pa gen vòt | medium | Standard | |
| 10 |
OpenVoice
Instant voice cloning with granular control over style, emotion, and accent.
300M
2024
|
4.0 /5 | Pa gen vòt | medium | Premium | |
| 11 |
IndexTTS-2
Zero-shot TTS with fine-grained emotion control and high expressiveness.
300M
2025
|
3.91 /5 | Pa gen vòt | medium | Standard | |
| 12 |
Spark TTS
Voice cloning TTS with controllable emotion and speaking style via prompts.
500M
2025
|
3.9 /5 | Pa gen vòt | medium | Standard | |
| 13 |
Parler TTS
Describe the voice you want in natural language and Parler generates matching speech.
880M
45000h
2024
|
3.83 /5 | Pa gen vòt | medium | Standard | |
| 14 |
Tortoise TTS
Multi-voice text-to-speech focused on quality with autoregressive architecture.
400M
50000h
2022
|
3.7 /5 | Pa gen vòt | slow | Premium | |
| 15 |
Bark
Transformer-based text-to-audio model that generates realistic speech, music, and sound effects.
350M
100000h
2023
|
3.57 /5 | Pa gen vòt | slow | Standard | |
| 16 |
Bark Small
Lighter version of Bark with faster inference and lower memory usage.
150M
100000h
2023
|
— | Pa gen vòt | medium | Standard | |
| 17 |
GLM-TTS
Achieves the lowest character error rate among open-source TTS models.
300M
2025
|
— | Pa gen vòt | medium | Standard | |
| 18 |
GPT-SoVITS
Few-shot voice cloning TTS that replicates any voice from just 5 seconds of audio.
200M
2024
|
— | Pa gen vòt | slow | Standard | |
| 19 |
Qwen3 TTS
Alibaba's multilingual TTS with voice cloning, preset voices, and voice design from text.
1.7B
2025
|
— | Pa gen vòt | medium | Standard | |
| 20 |
Sesame CSM
Conversational speech model generating natural dialogue with appropriate timing and emotion.
1B
2025
|
— | Pa gen vòt | slow | Premium | |
| 21 |
Chatterbox Turbo
Faster Chatterbox with sub-200ms latency and paralinguistic tags for laughs, coughs, and more.
350M
2025
|
— | Pa gen vòt | fast | Standard | |
| 22 |
Zonos
Emotion-controllable TTS with fine-grained sliders for happiness, anger, sadness, and more.
1.6B
200000h
2025
|
— | Pa gen vòt | medium | Standard | |
| 23 |
Dia 2
Streaming-first conversational TTS with multi-speaker dialogue and paralinguistic cues.
2B
2025
|
— | Pa gen vòt | fast | Standard | |
| 24 |
VoxCPM
Tokenizer-free TTS producing 44.1kHz audio with context-aware paragraph consistency.
500M
1800000h
2025
|
— | Pa gen vòt | fast | Standard | |
| 25 |
OuteTTS
LLM-based TTS that runs on CPU, GPU, or browser via llama.cpp and Transformers.js.
1B
5000h
2025
|
— | Pa gen vòt | fast | Free | |
| 26 |
TADA
Zero-hallucination TTS with text-acoustic dual alignment, 5x faster than comparable LLM TTS.
1B
2026
|
— | Pa gen vòt | fast | Standard | |
| 27 |
VibeVoice
Microsoft's multi-speaker long-form TTS generating up to 90 minutes with 4 distinct speakers.
1.5B
100000h
2025
|
— | Pa gen vòt | fast | Standard | |
| 28 |
Pocket TTS
Lightweight 100M parameter model by Kyutai with voice cloning from a single sample.
100M
50000h
2025
|
— | Pa gen vòt | fast | Free | |
| 29 |
Kitten TTS
Ultra-lightweight TTS under 80MB. Runs on CPU without GPU.
80M
2025
|
— | Pa gen vòt | fast | Free | |
| 30 |
CosyVoice3
Next-generation multilingual TTS with bi-streaming, emotion control, and zero-shot voice cloning.
500M
200000h
2025
|
— | Pa gen vòt | fast | Standard | |
| 31 |
MOSS-TTS
Ultra-long 20-language TTS supporting up to 1 hour of continuous generation with phoneme-level control.
8B
500000h
2026
|
— | Pa gen vòt | medium | Premium | |
| 32 |
MegaTTS3
ByteDance's sparse alignment TTS with adjustable intelligibility vs. speaker similarity.
1B
100000h
2025
|
— | Pa gen vòt | slow | Premium |
Resansman detaye
Ofisyèl TTS.ai benchmark pwen sou twa dimansyon: nati, presizyon, ak vitès.
Kokoro
Free
Natirèl
4.8/5
Presizyon
4.7/5
Vitès
4.9/5
An jeneral
4.8/5
CosyVoice 2
Standard
Natirèl
4.5/5
Presizyon
4.4/5
Vitès
3.8/5
An jeneral
4.26/5
Chatterbox
Premium
Natirèl
4.7/5
Presizyon
4.5/5
Vitès
3.4/5
An jeneral
4.25/5
StyleTTS 2
Premium
Natirèl
4.5/5
Presizyon
4.3/5
Vitès
3.8/5
An jeneral
4.23/5
Piper
Free
Natirèl
3.5/5
Presizyon
4.2/5
Vitès
4.95/5
An jeneral
4.15/5
MeloTTS
Free
Natirèl
3.8/5
Presizyon
4.1/5
Vitès
4.6/5
An jeneral
4.13/5
Dia TTS
Standard
Natirèl
4.6/5
Presizyon
4.3/5
Vitès
3.2/5
An jeneral
4.09/5
VITS
Free
Natirèl
3.4/5
Presizyon
4.0/5
Vitès
4.8/5
An jeneral
4.0/5
Orpheus
Standard
Natirèl
4.3/5
Presizyon
4.1/5
Vitès
3.5/5
An jeneral
4.0/5
OpenVoice
Premium
Natirèl
4.0/5
Presizyon
4.1/5
Vitès
3.9/5
An jeneral
4.0/5
IndexTTS-2
Standard
Natirèl
4.3/5
Presizyon
4.1/5
Vitès
3.2/5
An jeneral
3.91/5
Spark TTS
Standard
Natirèl
4.2/5
Presizyon
4.0/5
Vitès
3.4/5
An jeneral
3.9/5
Parler TTS
Standard
Natirèl
4.1/5
Presizyon
3.9/5
Vitès
3.4/5
An jeneral
3.83/5
Tortoise TTS
Premium
Natirèl
4.6/5
Presizyon
4.4/5
Vitès
1.8/5
An jeneral
3.7/5
Bark
Standard
Natirèl
4.2/5
Presizyon
3.8/5
Vitès
2.5/5
An jeneral
3.57/5
Metòd referans
Konfigurasyon Teste
- Machinè: 4x NVIDIA Tesla P40 (24GB VRAM chak), 96GB total
- Teste tèks: 5 pati estandardize k ap kouvri diferan modil pale (narrasyon, dyalòg, teknik, emosyonèl, multilengwistik)
- Evalyasyon: Metrik otomatik (estime MOS, WER, RTF) konbine ak tès tande moun
- Ekran: Chak modèl te teste 10 fwa pou chak pasaje, pwen mwayèn
Kritè pou Notasyon
- Natirèl (40%): Prozodi, intonasyon, rythme, emosyon — ki jan sa son pou yon moun?
- Presizyon (30%): Koreksyon pwononsyasyon, pousantaj erè mo, entelligibilite
- Vitès (30%): Faktori tan reyèl (segondè odyo / segondè jenerasyon). Pi wo = pi vit.
- An jeneral: Moyenne pondérée: 0.4 x Naturalness + 0.3 x Precision + 0.3 x Vitesse
Notifikasyon: Kalite nan mond reyèl la ka varye selon tèks enpòte, lang, ak seleksyon vwa. Rating kominote a bay yon siyal komplementè ki baze sou diferans nan itilize reyèl.
Kesyon ki poze souvan
What could we improve? Your feedback helps us fix issues.
Jwenn Voy ou a
Eseye nenpòt modèl gratis ak Kokoro, Piper, VITS, oswa MeloTTS. Pa gen kont mande.