TTS Arena — AI ahots modeloen sailkapena
Konparatu 20 testu-hizketa eredu baino gehiago. Erreferentzia ofizialak, komunitatearen balorazioak eta alboko konparazioa.
-Bai, bai.
Saldu zure ahotsa
Alboko konparaketa
Idatzi testua, hautatu bi eredu, eta konparatu emaitzak. Geruza libreko ereduek ez dute konturik behar.
Libre dauden modeloek konturik gabe funtzionatzen dute. Izena eman Premium modeloak konparatzeko.
Leaderboard modeloa
| # | Modeloa | Ofiziala | Komunitatea | Zure balorazioa | Abiadura | Animalia |
|---|---|---|---|---|---|---|
| 1 |
Kokoro
Lightweight 82M parameter model delivering studio-quality speech with blazing-fast inference.
82M
1200h
2024
|
4.8 /5 |
5.0
/5
1 botoa
|
fast | Free | |
| 2 |
CosyVoice 2
Alibaba's scalable streaming TTS with human-parity naturalness and near-zero latency.
300M
200000h
2024
|
4.26 /5 | Oraindik ez dago botorik | medium | Standard | |
| 3 |
Chatterbox
State-of-the-art zero-shot voice cloning with emotion control from Resemble AI.
300M
2025
|
4.25 /5 | Oraindik ez dago botorik | medium | Premium | |
| 4 |
StyleTTS 2
Human-level text-to-speech through style diffusion and adversarial training.
100M
585h
2024
|
4.23 /5 | Oraindik ez dago botorik | medium | Premium | |
| 5 |
Piper
A fast, local neural text to speech system optimized for Raspberry Pi and embedded devices.
15M
2023
|
4.15 /5 | Oraindik ez dago botorik | fast | Free | |
| 6 |
MeloTTS
High-quality multilingual text-to-speech that runs on CPU with minimal latency.
25M
2024
|
4.13 /5 | Oraindik ez dago botorik | fast | Free | |
| 7 |
Dia TTS
Multi-speaker dialog generation model that creates natural conversations between speakers.
1.6B
2024
|
4.09 /5 | Oraindik ez dago botorik | medium | Standard | |
| 8 |
VITS
Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech.
25M
585h
2021
|
4.0 /5 | Oraindik ez dago botorik | fast | Free | |
| 9 |
Orpheus
Human-level emotional TTS model trained on 100K hours of speech data.
3B
100000h
2025
|
4.0 /5 | Oraindik ez dago botorik | medium | Standard | |
| 10 |
OpenVoice
Instant voice cloning with granular control over style, emotion, and accent.
300M
2024
|
4.0 /5 | Oraindik ez dago botorik | medium | Premium | |
| 11 |
IndexTTS-2
Zero-shot TTS with fine-grained emotion control and high expressiveness.
300M
2025
|
3.91 /5 | Oraindik ez dago botorik | medium | Standard | |
| 12 |
Spark TTS
Voice cloning TTS with controllable emotion and speaking style via prompts.
500M
2025
|
3.9 /5 | Oraindik ez dago botorik | medium | Standard | |
| 13 |
Parler TTS
Describe the voice you want in natural language and Parler generates matching speech.
880M
45000h
2024
|
3.83 /5 | Oraindik ez dago botorik | medium | Standard | |
| 14 |
Tortoise TTS
Multi-voice text-to-speech focused on quality with autoregressive architecture.
400M
50000h
2022
|
3.7 /5 | Oraindik ez dago botorik | slow | Premium | |
| 15 |
Bark
Transformer-based text-to-audio model that generates realistic speech, music, and sound effects.
350M
100000h
2023
|
3.57 /5 | Oraindik ez dago botorik | slow | Standard | |
| 16 |
Bark Small
Lighter version of Bark with faster inference and lower memory usage.
150M
100000h
2023
|
— | Oraindik ez dago botorik | medium | Standard | |
| 17 |
GLM-TTS
Achieves the lowest character error rate among open-source TTS models.
300M
2025
|
— | Oraindik ez dago botorik | medium | Standard | |
| 18 |
GPT-SoVITS
Few-shot voice cloning TTS that replicates any voice from just 5 seconds of audio.
200M
2024
|
— | Oraindik ez dago botorik | slow | Standard | |
| 19 |
Qwen3 TTS
Alibaba's multilingual TTS with voice cloning, preset voices, and voice design from text.
1.7B
2025
|
— | Oraindik ez dago botorik | medium | Standard | |
| 20 |
Sesame CSM
Conversational speech model generating natural dialogue with appropriate timing and emotion.
1B
2025
|
— | Oraindik ez dago botorik | slow | Premium | |
| 21 |
Chatterbox Turbo
Faster Chatterbox with sub-200ms latency and paralinguistic tags for laughs, coughs, and more.
350M
2025
|
— | Oraindik ez dago botorik | fast | Standard | |
| 22 |
Zonos
Emotion-controllable TTS with fine-grained sliders for happiness, anger, sadness, and more.
1.6B
200000h
2025
|
— | Oraindik ez dago botorik | medium | Standard | |
| 23 |
Dia 2
Streaming-first conversational TTS with multi-speaker dialogue and paralinguistic cues.
2B
2025
|
— | Oraindik ez dago botorik | fast | Standard | |
| 24 |
VoxCPM
Tokenizer-free TTS producing 44.1kHz audio with context-aware paragraph consistency.
500M
1800000h
2025
|
— | Oraindik ez dago botorik | fast | Standard | |
| 25 |
OuteTTS
LLM-based TTS that runs on CPU, GPU, or browser via llama.cpp and Transformers.js.
1B
5000h
2025
|
— | Oraindik ez dago botorik | fast | Free | |
| 26 |
TADA
Zero-hallucination TTS with text-acoustic dual alignment, 5x faster than comparable LLM TTS.
1B
2026
|
— | Oraindik ez dago botorik | fast | Standard | |
| 27 |
VibeVoice
Microsoft's multi-speaker long-form TTS generating up to 90 minutes with 4 distinct speakers.
1.5B
100000h
2025
|
— | Oraindik ez dago botorik | fast | Standard | |
| 28 |
Pocket TTS
Lightweight 100M parameter model by Kyutai with voice cloning from a single sample.
100M
50000h
2025
|
— | Oraindik ez dago botorik | fast | Free | |
| 29 |
Kitten TTS
Ultra-lightweight TTS under 80MB. Runs on CPU without GPU.
80M
2025
|
— | Oraindik ez dago botorik | fast | Free | |
| 30 |
CosyVoice3
Next-generation multilingual TTS with bi-streaming, emotion control, and zero-shot voice cloning.
500M
200000h
2025
|
— | Oraindik ez dago botorik | 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
|
— | Oraindik ez dago botorik | medium | Premium | |
| 32 |
MegaTTS3
ByteDance's sparse alignment TTS with adjustable intelligibility vs. speaker similarity.
1B
100000h
2025
|
— | Oraindik ez dago botorik | slow | Premium |
Erreferentzia puntuazio xeheak
TTS.ai benchmark ofizialak hiru dimentsiotan puntuatzen du: naturaltasuna, zehaztasuna eta abiadura.
Kokoro
Free
Naturaltasuna
4.8/5
Zehaztasuna
4.7/5
Abiadura
4.9/5
Orokorra
4.8/5
CosyVoice 2
Standard
Naturaltasuna
4.5/5
Zehaztasuna
4.4/5
Abiadura
3.8/5
Orokorra
4.26/5
Chatterbox
Premium
Naturaltasuna
4.7/5
Zehaztasuna
4.5/5
Abiadura
3.4/5
Orokorra
4.25/5
StyleTTS 2
Premium
Naturaltasuna
4.5/5
Zehaztasuna
4.3/5
Abiadura
3.8/5
Orokorra
4.23/5
Piper
Free
Naturaltasuna
3.5/5
Zehaztasuna
4.2/5
Abiadura
4.95/5
Orokorra
4.15/5
MeloTTS
Free
Naturaltasuna
3.8/5
Zehaztasuna
4.1/5
Abiadura
4.6/5
Orokorra
4.13/5
Dia TTS
Standard
Naturaltasuna
4.6/5
Zehaztasuna
4.3/5
Abiadura
3.2/5
Orokorra
4.09/5
VITS
Free
Naturaltasuna
3.4/5
Zehaztasuna
4.0/5
Abiadura
4.8/5
Orokorra
4.0/5
Orpheus
Standard
Naturaltasuna
4.3/5
Zehaztasuna
4.1/5
Abiadura
3.5/5
Orokorra
4.0/5
OpenVoice
Premium
Naturaltasuna
4.0/5
Zehaztasuna
4.1/5
Abiadura
3.9/5
Orokorra
4.0/5
IndexTTS-2
Standard
Naturaltasuna
4.3/5
Zehaztasuna
4.1/5
Abiadura
3.2/5
Orokorra
3.91/5
Spark TTS
Standard
Naturaltasuna
4.2/5
Zehaztasuna
4.0/5
Abiadura
3.4/5
Orokorra
3.9/5
Parler TTS
Standard
Naturaltasuna
4.1/5
Zehaztasuna
3.9/5
Abiadura
3.4/5
Orokorra
3.83/5
Tortoise TTS
Premium
Naturaltasuna
4.6/5
Zehaztasuna
4.4/5
Abiadura
1.8/5
Orokorra
3.7/5
Bark
Standard
Naturaltasuna
4.2/5
Zehaztasuna
3.8/5
Abiadura
2.5/5
Orokorra
3.57/5
Erreferentzia-metodologia
Proba konfigurazioa
- Hardwarea: 4x NVIDIA Tesla P40 (24GB VRAM bakoitza), 96GB guztira
- Probako testua: 5 pasarte estandarizatu, hizketa-eredu desberdinak (narrazioa, elkarrizketa, teknikoa, emozionala, eleanitza)
- Ebaluazioa: Metrika automatikoak (MOS estimazioa, WER, RTF) giza entzumen-probekin konbinatuta
- Exekutatu: Eredu bakoitza 10 aldiz probatu da pasabide bakoitzean, puntuazioak batez bestekoak
Puntuazio-irizpideak
- Naturaltasuna (% 40): Prosodia, intonazioa, erritmoa, emozioa... zenbaterainoko gizatasuna du horrek?
- Zehaztasuna (% 30): Ahoskera zuzentasuna, hitzen errore tasa, ulergarritasuna
- Abiadura (% 30): Denbora errealeko faktorea (audio segundoak / sorkuntza segundoak). Handiagoa = azkarragoa.
- Orokorra: Batez besteko pisuduna: 0,4 x naturaltasuna + 0,3 x zehaztasuna + 0,3 x abiadura
Oharra: Benchmark-ek gure hardware eta proba-testu espezifikoen errendimendua islatzen dute. Benetako kalitatea aldakorra izan daiteke sarrerako testuaren, hizkuntzaren eta ahots-hautapenaren arabera. Komunitatearen balorazioek seinale osagarria ematen dute benetako erabileran oinarrituta.
Maiz egiten diren galderak
Zer hobetu dezakegu? Zure iritziak arazoak konpontzen laguntzen digu.
Aurkitu zure ahots perfektua
Probatu edozein modelo doan Kokoro, Piper, VITS edo MeloTTS-ekin. Ez da konturik behar.