TTS Arena LA TAKA A LA TURA Model Voice Liderboard
Compara els models de text de 20+ a veu. Els punts de referència oficials, les puntuacions de la comunitat i la comparació a costat.
Comparació de cara a cara
Escriviu text, escolliu dos models i compareu els resultats. Els models lliures no requereixen cap compte.
Els models lliures treballen sense cap compte. Signa comparar models de gràcia.
Millora del model
| # | Model | Oficial | Comunitat | La vostra puntuació | Velocitat | TierCity name (optional, probably does not need a translation) |
|---|---|---|---|---|---|---|
| 1 |
Kokoro
Lightweight 82M parameter model delivering studio-quality speech with blazing-fast inference.
82M
1200h
2024
|
4.8 /5 |
5.0
/5
1 votació
|
fast | Free | |
| 2 |
CosyVoice 2
Alibaba's scalable streaming TTS with human-parity naturalness and near-zero latency.
300M
200000h
2024
|
4.26 /5 | Encara no hi ha vots | medium | Standard | |
| 3 |
Chatterbox
State-of-the-art zero-shot voice cloning with emotion control from Resemble AI.
300M
2025
|
4.25 /5 | Encara no hi ha vots | medium | Premium | |
| 4 |
StyleTTS 2
Human-level text-to-speech through style diffusion and adversarial training.
100M
585h
2024
|
4.23 /5 | Encara no hi ha vots | medium | Premium | |
| 5 |
Piper
A fast, local neural text to speech system optimized for Raspberry Pi and embedded devices.
15M
2023
|
4.15 /5 | Encara no hi ha vots | fast | Free | |
| 6 |
MeloTTS
High-quality multilingual text-to-speech that runs on CPU with minimal latency.
25M
2024
|
4.13 /5 | Encara no hi ha vots | fast | Free | |
| 7 |
Dia TTS
Multi-speaker dialog generation model that creates natural conversations between speakers.
1.6B
2024
|
4.09 /5 | Encara no hi ha vots | medium | Standard | |
| 8 |
VITS
Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech.
25M
585h
2021
|
4.0 /5 | Encara no hi ha vots | fast | Free | |
| 9 |
Orpheus
Human-level emotional TTS model trained on 100K hours of speech data.
3B
100000h
2025
|
4.0 /5 | Encara no hi ha vots | medium | Standard | |
| 10 |
OpenVoice
Instant voice cloning with granular control over style, emotion, and accent.
300M
2024
|
4.0 /5 | Encara no hi ha vots | medium | Premium | |
| 11 |
IndexTTS-2
Zero-shot TTS with fine-grained emotion control and high expressiveness.
300M
2025
|
3.91 /5 | Encara no hi ha vots | medium | Standard | |
| 12 |
Spark TTS
Voice cloning TTS with controllable emotion and speaking style via prompts.
500M
2025
|
3.9 /5 | Encara no hi ha vots | medium | Standard | |
| 13 |
Parler TTS
Describe the voice you want in natural language and Parler generates matching speech.
880M
45000h
2024
|
3.83 /5 | Encara no hi ha vots | medium | Standard | |
| 14 |
Tortoise TTS
Multi-voice text-to-speech focused on quality with autoregressive architecture.
400M
50000h
2022
|
3.7 /5 | Encara no hi ha vots | slow | Premium | |
| 15 |
Bark
Transformer-based text-to-audio model that generates realistic speech, music, and sound effects.
350M
100000h
2023
|
3.57 /5 | Encara no hi ha vots | slow | Standard | |
| 16 |
Bark Small
Lighter version of Bark with faster inference and lower memory usage.
150M
100000h
2023
|
— | Encara no hi ha vots | medium | Standard | |
| 17 |
GLM-TTS
Achieves the lowest character error rate among open-source TTS models.
300M
2025
|
— | Encara no hi ha vots | medium | Standard | |
| 18 |
GPT-SoVITS
Few-shot voice cloning TTS that replicates any voice from just 5 seconds of audio.
200M
2024
|
— | Encara no hi ha vots | slow | Standard | |
| 19 |
Qwen3 TTS
Alibaba's multilingual TTS with voice cloning, preset voices, and voice design from text.
1.7B
2025
|
— | Encara no hi ha vots | medium | Standard | |
| 20 |
Sesame CSM
Conversational speech model generating natural dialogue with appropriate timing and emotion.
1B
2025
|
— | Encara no hi ha vots | slow | Premium | |
| 21 |
Chatterbox Turbo
Faster Chatterbox with sub-200ms latency and paralinguistic tags for laughs, coughs, and more.
350M
2025
|
— | Encara no hi ha vots | fast | Standard | |
| 22 |
Zonos
Emotion-controllable TTS with fine-grained sliders for happiness, anger, sadness, and more.
1.6B
200000h
2025
|
— | Encara no hi ha vots | medium | Standard | |
| 23 |
Dia 2
Streaming-first conversational TTS with multi-speaker dialogue and paralinguistic cues.
2B
2025
|
— | Encara no hi ha vots | fast | Standard | |
| 24 |
VoxCPM
Tokenizer-free TTS producing 44.1kHz audio with context-aware paragraph consistency.
500M
1800000h
2025
|
— | Encara no hi ha vots | fast | Standard | |
| 25 |
OuteTTS
LLM-based TTS that runs on CPU, GPU, or browser via llama.cpp and Transformers.js.
1B
5000h
2025
|
— | Encara no hi ha vots | fast | Free | |
| 26 |
TADA
Zero-hallucination TTS with text-acoustic dual alignment, 5x faster than comparable LLM TTS.
1B
2026
|
— | Encara no hi ha vots | fast | Standard | |
| 27 |
VibeVoice
Microsoft's multi-speaker long-form TTS generating up to 90 minutes with 4 distinct speakers.
1.5B
100000h
2025
|
— | Encara no hi ha vots | fast | Standard | |
| 28 |
Pocket TTS
Lightweight 100M parameter model by Kyutai with voice cloning from a single sample.
100M
50000h
2025
|
— | Encara no hi ha vots | fast | Free | |
| 29 |
Kitten TTS
Ultra-lightweight TTS under 80MB. Runs on CPU without GPU.
80M
2025
|
— | Encara no hi ha vots | fast | Free | |
| 30 |
CosyVoice3
Next-generation multilingual TTS with bi-streaming, emotion control, and zero-shot voice cloning.
500M
200000h
2025
|
— | Encara no hi ha vots | 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
|
— | Encara no hi ha vots | medium | Premium | |
| 32 |
MegaTTS3
ByteDance's sparse alignment TTS with adjustable intelligibility vs. speaker similarity.
1B
100000h
2025
|
— | Encara no hi ha vots | slow | Premium |
Puntuació detallada de Benchmark
Official TTS.ai benchmark scores across three dimensions: naturalness, accuracy, and speed.
Kokoro
Free
CosyVoice 2
Standard
Chatterbox
Premium
StyleTTS 2
Premium
Piper
Free
MeloTTS
Free
Dia TTS
Standard
VITS
Free
Orpheus
Standard
OpenVoice
Premium
IndexTTS-2
Standard
Spark TTS
Standard
Parler TTS
Standard
Tortoise TTS
Premium
Bark
Standard
Metodologia de Benchmark
Configuració de la prova
- Maquinari: 4x NVIDIA Tesla P40 (24GB VRA), 96GB total
- Text de prova: Cinc passatges estàndard que cobreixen diferents patrons de parla (naració, diàleg, tècnica, emocional i multilingüe)
- avaluació: mètriques automatitzades (ordre KADMOS, WOR, RTF) combinades amb les proves d' escolta humana
- Executa: Cada model prova 10 vegades per pas, puntua la mitjana
Criteri de puntuació
- Naturalitat (40%): Prosodi, inneció, ritme, emoció, com sona l'humà?
- Exactitud (30%): Pronunciació correcta, taxa d' error de paraula, interelligibilitat
- Velocitat (30%): Factor de temps real (audi segons / generació). Més alt = més ràpid.
- En general: Mitjana ponderada: 0. 4 x Naturalitat + 0. x = 0. 3 x = 0. 3 x Velocitat
Nota: Benchmarks reflecteixen el rendiment del nostre maquinari específic i dels textos de prova. La qualitat real del món pot variar basant- se en el text d' entrada, la llengua i la selecció de veu. Les puntuacions comunitàries proporcionen un senyal complementària basat en un ús divers.
Preguntes més freqüents
What could we improve? Your feedback helps us fix issues.
Trobeu la vostra veu perfecta
Prova qualsevol model lliure amb Kokoro, Piper, VITS o MeloTTS.