TTS Arena — Laharana ambony amin'ny modely feon'ny AI
Ampitahao ny maodely 20+ amin'ny fandikana lahatsoratra ho teny. ny mari-pamantarana ofisialy, ny mari-pankasitrahan'ny vondrom-piarahamonina, ary ny fampitahana an-tsehatra.
Tsy mbola manana feo TTS amin'ny teny malagasy izahay. Ataovy izay hampidirina ny anareo!
Mivarotra ny Feonao
Fampitahana sisiny amin'ny sisiny
Manorata lahabolana, misafidy modely roa, ary ampitahao ny vokatra. Tsy mila kaonty ny modely maimaimpoana.
Ireo modely maimaimpoana dia miasa tsy misy kaonty. Misoratra anarana mba hampitaha ireo modely premium.
Modelin'ny Leaderboard
| # | Modely | Oficiel | Fiarahamonina | Ny mari-pankasitrahanao | Hafainganan'ny fanovana | Taona |
|---|---|---|---|---|---|---|
| 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 | Tsy mbola misy latsa-bato | medium | Standard | |
| 3 |
Chatterbox
State-of-the-art zero-shot voice cloning with emotion control from Resemble AI.
300M
2025
|
4.25 /5 | Tsy mbola misy latsa-bato | medium | Premium | |
| 4 |
StyleTTS 2
Human-level text-to-speech through style diffusion and adversarial training.
100M
585h
2024
|
4.23 /5 | Tsy mbola misy latsa-bato | medium | Premium | |
| 5 |
Piper
A fast, local neural text to speech system optimized for Raspberry Pi and embedded devices.
15M
2023
|
4.15 /5 | Tsy mbola misy latsa-bato | fast | Free | |
| 6 |
MeloTTS
High-quality multilingual text-to-speech that runs on CPU with minimal latency.
25M
2024
|
4.13 /5 | Tsy mbola misy latsa-bato | fast | Free | |
| 7 |
Dia TTS
Multi-speaker dialog generation model that creates natural conversations between speakers.
1.6B
2024
|
4.09 /5 | Tsy mbola misy latsa-bato | medium | Standard | |
| 8 |
VITS
Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech.
25M
585h
2021
|
4.0 /5 | Tsy mbola misy latsa-bato | fast | Free | |
| 9 |
Orpheus
Human-level emotional TTS model trained on 100K hours of speech data.
3B
100000h
2025
|
4.0 /5 | Tsy mbola misy latsa-bato | medium | Standard | |
| 10 |
OpenVoice
Instant voice cloning with granular control over style, emotion, and accent.
300M
2024
|
4.0 /5 | Tsy mbola misy latsa-bato | medium | Premium | |
| 11 |
IndexTTS-2
Zero-shot TTS with fine-grained emotion control and high expressiveness.
300M
2025
|
3.91 /5 | Tsy mbola misy latsa-bato | medium | Standard | |
| 12 |
Spark TTS
Voice cloning TTS with controllable emotion and speaking style via prompts.
500M
2025
|
3.9 /5 | Tsy mbola misy latsa-bato | medium | Standard | |
| 13 |
Parler TTS
Describe the voice you want in natural language and Parler generates matching speech.
880M
45000h
2024
|
3.83 /5 | Tsy mbola misy latsa-bato | medium | Standard | |
| 14 |
Tortoise TTS
Multi-voice text-to-speech focused on quality with autoregressive architecture.
400M
50000h
2022
|
3.7 /5 | Tsy mbola misy latsa-bato | slow | Premium | |
| 15 |
Bark
Transformer-based text-to-audio model that generates realistic speech, music, and sound effects.
350M
100000h
2023
|
3.57 /5 | Tsy mbola misy latsa-bato | slow | Standard | |
| 16 |
Bark Small
Lighter version of Bark with faster inference and lower memory usage.
150M
100000h
2023
|
— | Tsy mbola misy latsa-bato | medium | Standard | |
| 17 |
GLM-TTS
Achieves the lowest character error rate among open-source TTS models.
300M
2025
|
— | Tsy mbola misy latsa-bato | medium | Standard | |
| 18 |
GPT-SoVITS
Few-shot voice cloning TTS that replicates any voice from just 5 seconds of audio.
200M
2024
|
— | Tsy mbola misy latsa-bato | slow | Standard | |
| 19 |
Qwen3 TTS
Alibaba's multilingual TTS with voice cloning, preset voices, and voice design from text.
1.7B
2025
|
— | Tsy mbola misy latsa-bato | medium | Standard | |
| 20 |
Sesame CSM
Conversational speech model generating natural dialogue with appropriate timing and emotion.
1B
2025
|
— | Tsy mbola misy latsa-bato | slow | Premium | |
| 21 |
Chatterbox Turbo
Faster Chatterbox with sub-200ms latency and paralinguistic tags for laughs, coughs, and more.
350M
2025
|
— | Tsy mbola misy latsa-bato | fast | Standard | |
| 22 |
Zonos
Emotion-controllable TTS with fine-grained sliders for happiness, anger, sadness, and more.
1.6B
200000h
2025
|
— | Tsy mbola misy latsa-bato | medium | Standard | |
| 23 |
Dia 2
Streaming-first conversational TTS with multi-speaker dialogue and paralinguistic cues.
2B
2025
|
— | Tsy mbola misy latsa-bato | fast | Standard | |
| 24 |
VoxCPM
Tokenizer-free TTS producing 44.1kHz audio with context-aware paragraph consistency.
500M
1800000h
2025
|
— | Tsy mbola misy latsa-bato | fast | Standard | |
| 25 |
OuteTTS
LLM-based TTS that runs on CPU, GPU, or browser via llama.cpp and Transformers.js.
1B
5000h
2025
|
— | Tsy mbola misy latsa-bato | fast | Free | |
| 26 |
TADA
Zero-hallucination TTS with text-acoustic dual alignment, 5x faster than comparable LLM TTS.
1B
2026
|
— | Tsy mbola misy latsa-bato | fast | Standard | |
| 27 |
VibeVoice
Microsoft's multi-speaker long-form TTS generating up to 90 minutes with 4 distinct speakers.
1.5B
100000h
2025
|
— | Tsy mbola misy latsa-bato | fast | Standard | |
| 28 |
Pocket TTS
Lightweight 100M parameter model by Kyutai with voice cloning from a single sample.
100M
50000h
2025
|
— | Tsy mbola misy latsa-bato | fast | Free | |
| 29 |
Kitten TTS
Ultra-lightweight TTS under 80MB. Runs on CPU without GPU.
80M
2025
|
— | Tsy mbola misy latsa-bato | fast | Free | |
| 30 |
CosyVoice3
Next-generation multilingual TTS with bi-streaming, emotion control, and zero-shot voice cloning.
500M
200000h
2025
|
— | Tsy mbola misy latsa-bato | 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
|
— | Tsy mbola misy latsa-bato | medium | Premium | |
| 32 |
MegaTTS3
ByteDance's sparse alignment TTS with adjustable intelligibility vs. speaker similarity.
1B
100000h
2025
|
— | Tsy mbola misy latsa-bato | slow | Premium |
Toetry ny benchmark amin'ny antsipiriany
Ny TTS.ai dia manome isa amin'ny lafiny telo: ny natiora, ny fahamarinana, ary ny hafainganam-pandeha.
Kokoro
Free
Natiora
4.8/5
Fahasamihafana
4.7/5
_Hafainganana:
4.9/5
Ankapobeny
4.8/5
CosyVoice 2
Standard
Natiora
4.5/5
Fahasamihafana
4.4/5
_Hafainganana:
3.8/5
Ankapobeny
4.26/5
Chatterbox
Premium
Natiora
4.7/5
Fahasamihafana
4.5/5
_Hafainganana:
3.4/5
Ankapobeny
4.25/5
StyleTTS 2
Premium
Natiora
4.5/5
Fahasamihafana
4.3/5
_Hafainganana:
3.8/5
Ankapobeny
4.23/5
Piper
Free
Natiora
3.5/5
Fahasamihafana
4.2/5
_Hafainganana:
4.95/5
Ankapobeny
4.15/5
MeloTTS
Free
Natiora
3.8/5
Fahasamihafana
4.1/5
_Hafainganana:
4.6/5
Ankapobeny
4.13/5
Dia TTS
Standard
Natiora
4.6/5
Fahasamihafana
4.3/5
_Hafainganana:
3.2/5
Ankapobeny
4.09/5
VITS
Free
Natiora
3.4/5
Fahasamihafana
4.0/5
_Hafainganana:
4.8/5
Ankapobeny
4.0/5
Orpheus
Standard
Natiora
4.3/5
Fahasamihafana
4.1/5
_Hafainganana:
3.5/5
Ankapobeny
4.0/5
OpenVoice
Premium
Natiora
4.0/5
Fahasamihafana
4.1/5
_Hafainganana:
3.9/5
Ankapobeny
4.0/5
IndexTTS-2
Standard
Natiora
4.3/5
Fahasamihafana
4.1/5
_Hafainganana:
3.2/5
Ankapobeny
3.91/5
Spark TTS
Standard
Natiora
4.2/5
Fahasamihafana
4.0/5
_Hafainganana:
3.4/5
Ankapobeny
3.9/5
Parler TTS
Standard
Natiora
4.1/5
Fahasamihafana
3.9/5
_Hafainganana:
3.4/5
Ankapobeny
3.83/5
Tortoise TTS
Premium
Natiora
4.6/5
Fahasamihafana
4.4/5
_Hafainganana:
1.8/5
Ankapobeny
3.7/5
Bark
Standard
Natiora
4.2/5
Fahasamihafana
3.8/5
_Hafainganana:
2.5/5
Ankapobeny
3.57/5
Fitaovana
Fikirakirana fitsapana
- Fitaovana: 4x NVIDIA Tesla P40 (24GB VRAM isaky ny iray), 96GB ny totaliny
- Soratra fitsapana: 5 ny andininy voasoratra izay mirakitra ny fomba fiteny samihafa (fitantarana, fifanakalozan-kevitra, teknika, fihetseham-po, fiteny maro)
- Fandinihana: Famaritana mandeha ho azy (fanadihadiana MOS, WER, RTF) miaraka amin'ny fitsapana ny fihainoan'ny olona
- Mandeha: Ny modely tsirairay dia voamarina in-10 isaky ny dingana, ny salan'isan'ny isa
Famaritana ny isa
- Natiora (40%): Ny fiteny, ny fiteny, ny hira, ny fihetseham-po — ahoana no maha-olona azy?
- Fahasamihafana (30%): Ny fahamarinan'ny fehezanteny, ny tahan'ny tsy fetezan'ny teny, ny fahamendrehana
- Hafainganana (30%): Isan'ny fotoana tena izy (segondra amin'ny feo / segondra amin'ny famoronana). Avo kokoa = haingana kokoa.
- Ankapobeny: Isam-pahefana: 0.4 x Natiora + 0.3 x Famaritana + 0.3 x Hafainganana
Fanamarihana: Maneho ny fahombiazana amin'ny fitaovana manokana sy ny lahatsoratra fanandramana ny mari-pamantarana. Mety miovaova arakaraka ny lahabolana, ny teny, ary ny safidin'ny feo ny kalitaon'ny tontolo tena izy. Manome famantarana fanampiana mifototra amin'ny fampiasana tena misy ny mari-pankasitrahan'ny vondrom-piarahamonina.
Fanontaniana mipetraka matetika
Inona no azonay atao mba hanatsarana? Manampy anay hamahana ny olana ny fihetseham-ponao.
Mitadiava ny Feon'ny Tenao
Andramo maimaimpoana ny modely amin'ny Kokoro, Piper, VITS, na MeloTTS. Tsy mila kaonty.