TTS Arena — AI Voice Model Leaderboard
Akkasumas, 20+ teeksta-hiika-dhaamsa. Benchmarks, fuula-gaaffilee, fi akkasumas-gaaffilee.
Haata'u malee, afaan keessaniin dhageettiwwan TTS hin qabnu. Jaalchiina keenyatti addeessi!
Dhaamsa keessan Sell
Akkasumas
Haala tokko galchi, mo'ellaa lamaan fila, fi itti gaafatama walqabatee. Mo'ellaa-tarree-free akkasii hin barbaachisu.
Moodeeloonni hin-barbaachisin hojiirra oolchuu hin dandeenye. Jijjiirama to compare premium models.
Model Leaderboard
| # | Modelii | Ofiishaa | Community | Rakkoo keessan | Jijjiiramni | Daandiin |
|---|---|---|---|---|---|---|
| 1 |
Kokoro
Lightweight 82M parameter model delivering studio-quality speech with blazing-fast inference.
82M
1200h
2024
|
4.8 /5 |
5.0
/5
1 fuula
|
fast | Free | |
| 2 |
CosyVoice 2
Alibaba's scalable streaming TTS with human-parity naturalness and near-zero latency.
300M
200000h
2024
|
4.26 /5 | Haata'u malee filannoo hin jiru | medium | Standard | |
| 3 |
Chatterbox
State-of-the-art zero-shot voice cloning with emotion control from Resemble AI.
300M
2025
|
4.25 /5 | Haata'u malee filannoo hin jiru | medium | Premium | |
| 4 |
StyleTTS 2
Human-level text-to-speech through style diffusion and adversarial training.
100M
585h
2024
|
4.23 /5 | Haata'u malee filannoo hin jiru | medium | Premium | |
| 5 |
Piper
A fast, local neural text to speech system optimized for Raspberry Pi and embedded devices.
15M
2023
|
4.15 /5 | Haata'u malee filannoo hin jiru | fast | Free | |
| 6 |
MeloTTS
High-quality multilingual text-to-speech that runs on CPU with minimal latency.
25M
2024
|
4.13 /5 | Haata'u malee filannoo hin jiru | fast | Free | |
| 7 |
Dia TTS
Multi-speaker dialog generation model that creates natural conversations between speakers.
1.6B
2024
|
4.09 /5 | Haata'u malee filannoo hin jiru | medium | Standard | |
| 8 |
VITS
Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech.
25M
585h
2021
|
4.0 /5 | Haata'u malee filannoo hin jiru | fast | Free | |
| 9 |
Orpheus
Human-level emotional TTS model trained on 100K hours of speech data.
3B
100000h
2025
|
4.0 /5 | Haata'u malee filannoo hin jiru | medium | Standard | |
| 10 |
OpenVoice
Instant voice cloning with granular control over style, emotion, and accent.
300M
2024
|
4.0 /5 | Haata'u malee filannoo hin jiru | medium | Premium | |
| 11 |
IndexTTS-2
Zero-shot TTS with fine-grained emotion control and high expressiveness.
300M
2025
|
3.91 /5 | Haata'u malee filannoo hin jiru | medium | Standard | |
| 12 |
Spark TTS
Voice cloning TTS with controllable emotion and speaking style via prompts.
500M
2025
|
3.9 /5 | Haata'u malee filannoo hin jiru | medium | Standard | |
| 13 |
Parler TTS
Describe the voice you want in natural language and Parler generates matching speech.
880M
45000h
2024
|
3.83 /5 | Haata'u malee filannoo hin jiru | medium | Standard | |
| 14 |
Tortoise TTS
Multi-voice text-to-speech focused on quality with autoregressive architecture.
400M
50000h
2022
|
3.7 /5 | Haata'u malee filannoo hin jiru | slow | Premium | |
| 15 |
Bark
Transformer-based text-to-audio model that generates realistic speech, music, and sound effects.
350M
100000h
2023
|
3.57 /5 | Haata'u malee filannoo hin jiru | slow | Standard | |
| 16 |
Bark Small
Lighter version of Bark with faster inference and lower memory usage.
150M
100000h
2023
|
— | Haata'u malee filannoo hin jiru | medium | Standard | |
| 17 |
Indic Parler TTS
High-quality speech for 8+ Indian languages with natural-language voice control.
900M
8000h
2024
|
— | Haata'u malee filannoo hin jiru | slow | Standard | |
| 18 |
KhanomTan TTS
Thai-first text-to-speech with a choice of speaker voices.
85M
100h
2023
|
— | Haata'u malee filannoo hin jiru | fast | Standard | |
| 19 |
GPT-SoVITS
Few-shot voice cloning TTS that replicates any voice from just 5 seconds of audio.
200M
2024
|
— | Haata'u malee filannoo hin jiru | slow | Standard | |
| 20 |
Qwen3 TTS
Alibaba's multilingual TTS with preset voices and voice design from text.
1.7B
2025
|
— | Haata'u malee filannoo hin jiru | medium | Standard | |
| 21 |
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
|
— | Haata'u malee filannoo hin jiru | fast | Standard | |
| 22 |
Sesame CSM
Conversational speech model generating natural dialogue with appropriate timing and emotion.
1B
2025
|
— | Haata'u malee filannoo hin jiru | slow | Premium | |
| 23 |
Chatterbox Turbo
Faster Chatterbox with sub-200ms latency and paralinguistic tags for laughs, coughs, and more.
350M
2025
|
— | Haata'u malee filannoo hin jiru | fast | Standard | |
| 24 |
VoxCPM
Tokenizer-free TTS producing 44.1kHz audio with context-aware paragraph consistency.
500M
1800000h
2025
|
— | Haata'u malee filannoo hin jiru | fast | Standard | |
| 25 |
Kani TTS 2
Ultra-lightweight 400M English TTS model running in just 3GB VRAM.
400M
10000h
2026
|
— | Haata'u malee filannoo hin jiru | fast | Free | |
| 26 |
OuteTTS
LLM-based TTS that runs on CPU, GPU, or browser via llama.cpp and Transformers.js.
1B
5000h
2025
|
— | Haata'u malee filannoo hin jiru | slow | Free | |
| 27 |
VibeVoice
Microsoft's multi-speaker long-form TTS generating up to 90 minutes with 4 distinct speakers.
1.5B
100000h
2025
|
— | Haata'u malee filannoo hin jiru | fast | Standard | |
| 28 |
Pocket TTS
Lightweight 100M parameter model by Kyutai with voice cloning from a single sample.
100M
50000h
2025
|
— | Haata'u malee filannoo hin jiru | fast | Free | |
| 29 |
Kitten TTS
Ultra-lightweight TTS under 80MB. Runs on CPU without GPU.
80M
2025
|
— | Haata'u malee filannoo hin jiru | fast | Free | |
| 30 |
CosyVoice3
Next-generation multilingual TTS with bi-streaming, emotion control, and zero-shot voice cloning.
500M
200000h
2025
|
— | Haata'u malee filannoo hin jiru | fast | Standard | |
| 31 |
NAMAA Saudi TTS
First open Saudi-Arabic TTS. Native Saudi dialect with Chatterbox-quality voice cloning.
300M
2026
|
— | Haata'u malee filannoo hin jiru | medium | Standard | |
| 32 |
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
|
— | Haata'u malee filannoo hin jiru | medium | Standard | |
| 33 |
MOSS-TTSD
Multi-speaker dialogue continuation model — generate podcast-style conversations with up to 5 speakers and 60 minutes of coherent audio.
7B
2026
|
— | Haata'u malee filannoo hin jiru | medium | Standard | |
| 34 |
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
|
— | Haata'u malee filannoo hin jiru | medium | Free | |
| 35 |
MOSS-TTS Nano
Tiny 100M MOSS-TTS variant — same architecture, 80x smaller, free-tier latency.
100M
500000h
2026
|
— | Haata'u malee filannoo hin jiru | fast | Free |
Fuula
TTS.ai Benchmark fuula fuula tokkoon tokkoon isaa: naanno, fuula, fi jabaa.
Kokoro
Free
Nafxanyaa
4.8/5
Akkasumas
4.7/5
Jijjiiramni
4.9/5
Fuula
4.8/5
CosyVoice 2
Standard
Nafxanyaa
4.5/5
Akkasumas
4.4/5
Jijjiiramni
3.8/5
Fuula
4.26/5
Chatterbox
Premium
Nafxanyaa
4.7/5
Akkasumas
4.5/5
Jijjiiramni
3.4/5
Fuula
4.25/5
StyleTTS 2
Premium
Nafxanyaa
4.5/5
Akkasumas
4.3/5
Jijjiiramni
3.8/5
Fuula
4.23/5
Piper
Free
Nafxanyaa
3.5/5
Akkasumas
4.2/5
Jijjiiramni
4.95/5
Fuula
4.15/5
MeloTTS
Free
Nafxanyaa
3.8/5
Akkasumas
4.1/5
Jijjiiramni
4.6/5
Fuula
4.13/5
Dia TTS
Standard
Nafxanyaa
4.6/5
Akkasumas
4.3/5
Jijjiiramni
3.2/5
Fuula
4.09/5
VITS
Free
Nafxanyaa
3.4/5
Akkasumas
4.0/5
Jijjiiramni
4.8/5
Fuula
4.0/5
Orpheus
Standard
Nafxanyaa
4.3/5
Akkasumas
4.1/5
Jijjiiramni
3.5/5
Fuula
4.0/5
OpenVoice
Premium
Nafxanyaa
4.0/5
Akkasumas
4.1/5
Jijjiiramni
3.9/5
Fuula
4.0/5
IndexTTS-2
Standard
Nafxanyaa
4.3/5
Akkasumas
4.1/5
Jijjiiramni
3.2/5
Fuula
3.91/5
Spark TTS
Standard
Nafxanyaa
4.2/5
Akkasumas
4.0/5
Jijjiiramni
3.4/5
Fuula
3.9/5
Parler TTS
Standard
Nafxanyaa
4.1/5
Akkasumas
3.9/5
Jijjiiramni
3.4/5
Fuula
3.83/5
Tortoise TTS
Premium
Nafxanyaa
4.6/5
Akkasumas
4.4/5
Jijjiiramni
1.8/5
Fuula
3.7/5
Bark
Standard
Nafxanyaa
4.2/5
Akkasumas
3.8/5
Jijjiiramni
2.5/5
Fuula
3.57/5
Me'toodooloojii
Akkaataa Jijjiiramaa
- Haaraa: 4x NVIDIA Tesla P40 (24GB VRAM tokko tokko), 96GB guutuu
- Akkaata qorannoo: 5-gaaffilee fuula fuula 5-gaaffilee fuula fuula 5-gaaffilee fuula 5-gaaffilee fuula 5-gaaffilee fuula 5-gaaffilee fuula 5-gaaffilee fuula 5
- Filannoo: Metrics Automated (MOS estimation, WER, RTF) walfakkaatuun fi qorannoo dhagahuu namaa
- Ji'oota: Modeelli tokko tokko yeroo 10 dura itti fayyadama, galmeewwan baay'ee
Filannoo
- Nafxanyaa (40%): Prosodia, intonaansi, riim, emoashion—maaliif akka namaatti dhagahama?
- Akkasumas (30%): Qabeenya barreessuu, dhibee galattoota, fahmu
- Jijjiiramni (30%): Fakkii yeroo-dhabdee (sekondii awdiyoo / sekondii ji'iinsaa). Garaa = jabaa.
- Akkasumas: Akkasumas: 0.4 x Nafxanyaa + 0.3 x Akkasumas + 0.3 x Ibsa
Fakkeenya: Benchmarks'n hojiirra oolchuu irratti haadha ruuraa fi kitaaba nuti ittisaa jirru mul'isa. Qunnamtii addunyaa dhabda'a kan wal fakkaatu yoo ta'e kan akka kitaaba, afaanii, fi filannoo dhalootaaf. Rakkoota hawaasaa siinii komplementaariif ta'e yoo ta'e kan akka fayyadama dhabda'a.
Su'aalota yeroo dheeraaf dhiyaatan
Maaliif nu barbaachisa? Dhugaa kee nu gargaara rakkoolee ittisaa.
Fuula kee guutuu
Akkasumas, Kokoro, Piper, VITS, fi MeloTTS fayyadamuun, modelaa tokko malee fayyadami. Akkasumas, akkasii hin barbaachisu.