TTS Arena! AI Voice Model Leaderboard
Sammenlign 20+ tekst-til-tale modeller. Officielle benchmarks, community ratings, og side-by-side sammenligning.
Sammenligning side for side
Skriv tekst, vælg to modeller, og sammenlign resultaterne. Gratis modeller kræver ingen konto.
Gratis modeller fungerer uden en konto. Tilmeld dig at sammenligne premium modeller.
Model Leaderboard
| # | Model | De Europæiske Fællesskabers Tidende | Fællesskabet | Din bedømmelse | Hastighed | Metodetrin |
|---|---|---|---|---|---|---|
| 1 |
Kokoro
Lightweight 82M parameter model delivering studio-quality speech with blazing-fast inference.
82M
1200h
2024
|
4.8 /5 |
5.0
/5
1 Afstemning
|
fast | Free | |
| 2 |
CosyVoice 2
Alibaba's scalable streaming TTS with human-parity naturalness and near-zero latency.
300M
200000h
2024
|
4.26 /5 | Ingen stemmer endnu | medium | Standard | |
| 3 |
Chatterbox
State-of-the-art zero-shot voice cloning with emotion control from Resemble AI.
300M
2025
|
4.25 /5 | Ingen stemmer endnu | medium | Premium | |
| 4 |
StyleTTS 2
Human-level text-to-speech through style diffusion and adversarial training.
100M
585h
2024
|
4.23 /5 | Ingen stemmer endnu | medium | Premium | |
| 5 |
Piper
A fast, local neural text to speech system optimized for Raspberry Pi and embedded devices.
15M
2023
|
4.15 /5 | Ingen stemmer endnu | fast | Free | |
| 6 |
MeloTTS
High-quality multilingual text-to-speech that runs on CPU with minimal latency.
25M
2024
|
4.13 /5 | Ingen stemmer endnu | fast | Free | |
| 7 |
Dia TTS
Multi-speaker dialog generation model that creates natural conversations between speakers.
1.6B
2024
|
4.09 /5 | Ingen stemmer endnu | medium | Standard | |
| 8 |
VITS
Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech.
25M
585h
2021
|
4.0 /5 | Ingen stemmer endnu | fast | Free | |
| 9 |
Orpheus
Human-level emotional TTS model trained on 100K hours of speech data.
3B
100000h
2025
|
4.0 /5 | Ingen stemmer endnu | medium | Standard | |
| 10 |
OpenVoice
Instant voice cloning with granular control over style, emotion, and accent.
300M
2024
|
4.0 /5 | Ingen stemmer endnu | medium | Premium | |
| 11 |
IndexTTS-2
Zero-shot TTS with fine-grained emotion control and high expressiveness.
300M
2025
|
3.91 /5 | Ingen stemmer endnu | medium | Standard | |
| 12 |
Spark TTS
Voice cloning TTS with controllable emotion and speaking style via prompts.
500M
2025
|
3.9 /5 | Ingen stemmer endnu | medium | Standard | |
| 13 |
Parler TTS
Describe the voice you want in natural language and Parler generates matching speech.
880M
45000h
2024
|
3.83 /5 | Ingen stemmer endnu | medium | Standard | |
| 14 |
Tortoise TTS
Multi-voice text-to-speech focused on quality with autoregressive architecture.
400M
50000h
2022
|
3.7 /5 | Ingen stemmer endnu | slow | Premium | |
| 15 |
Bark
Transformer-based text-to-audio model that generates realistic speech, music, and sound effects.
350M
100000h
2023
|
3.57 /5 | Ingen stemmer endnu | slow | Standard | |
| 16 |
Bark Small
Lighter version of Bark with faster inference and lower memory usage.
150M
100000h
2023
|
— | Ingen stemmer endnu | medium | Standard | |
| 17 |
GLM-TTS
Achieves the lowest character error rate among open-source TTS models.
300M
2025
|
— | Ingen stemmer endnu | medium | Standard | |
| 18 |
GPT-SoVITS
Few-shot voice cloning TTS that replicates any voice from just 5 seconds of audio.
200M
2024
|
— | Ingen stemmer endnu | slow | Standard | |
| 19 |
Qwen3 TTS
Alibaba's multilingual TTS with voice cloning, preset voices, and voice design from text.
1.7B
2025
|
— | Ingen stemmer endnu | medium | Standard | |
| 20 |
Sesame CSM
Conversational speech model generating natural dialogue with appropriate timing and emotion.
1B
2025
|
— | Ingen stemmer endnu | slow | Premium | |
| 21 |
Chatterbox Turbo
Faster Chatterbox with sub-200ms latency and paralinguistic tags for laughs, coughs, and more.
350M
2025
|
— | Ingen stemmer endnu | fast | Standard | |
| 22 |
Dia 2
Streaming-first conversational TTS with multi-speaker dialogue and paralinguistic cues.
2B
2025
|
— | Ingen stemmer endnu | fast | Standard | |
| 23 |
VoxCPM
Tokenizer-free TTS producing 44.1kHz audio with context-aware paragraph consistency.
500M
1800000h
2025
|
— | Ingen stemmer endnu | fast | Standard | |
| 24 |
OuteTTS
LLM-based TTS that runs on CPU, GPU, or browser via llama.cpp and Transformers.js.
1B
5000h
2025
|
— | Ingen stemmer endnu | fast | Free | |
| 25 |
TADA
Zero-hallucination TTS with text-acoustic dual alignment, 5x faster than comparable LLM TTS.
1B
2026
|
— | Ingen stemmer endnu | fast | Standard | |
| 26 |
VibeVoice
Microsoft's multi-speaker long-form TTS generating up to 90 minutes with 4 distinct speakers.
1.5B
100000h
2025
|
— | Ingen stemmer endnu | fast | Standard | |
| 27 |
Pocket TTS
Lightweight 100M parameter model by Kyutai with voice cloning from a single sample.
100M
50000h
2025
|
— | Ingen stemmer endnu | fast | Free | |
| 28 |
Kitten TTS
Ultra-lightweight TTS under 80MB. Runs on CPU without GPU.
80M
2025
|
— | Ingen stemmer endnu | fast | Free | |
| 29 |
CosyVoice3
Next-generation multilingual TTS with bi-streaming, emotion control, and zero-shot voice cloning.
500M
200000h
2025
|
— | Ingen stemmer endnu | fast | Standard | |
| 30 |
MOSS-TTS
Ultra-long 20-language TTS supporting up to 1 hour of continuous generation with phoneme-level control.
8B
500000h
2026
|
— | Ingen stemmer endnu | medium | Premium | |
| 31 |
MegaTTS3
ByteDance's sparse alignment TTS with adjustable intelligibility vs. speaker similarity.
1B
100000h
2025
|
— | Ingen stemmer endnu | slow | Premium |
Detaljerede benchmarkscore
Officiel TTS.ai benchmark scorer på tværs af tre dimensioner: naturlighed, nøjagtighed og hastighed.
Kokoro
Free
Naturlighed
4.8/5
Nøjagtighed
4.7/5
Hastighed
4.9/5
I alt
4.8/5
CosyVoice 2
Standard
Naturlighed
4.5/5
Nøjagtighed
4.4/5
Hastighed
3.8/5
I alt
4.26/5
Chatterbox
Premium
Naturlighed
4.7/5
Nøjagtighed
4.5/5
Hastighed
3.4/5
I alt
4.25/5
StyleTTS 2
Premium
Naturlighed
4.5/5
Nøjagtighed
4.3/5
Hastighed
3.8/5
I alt
4.23/5
Piper
Free
Naturlighed
3.5/5
Nøjagtighed
4.2/5
Hastighed
4.95/5
I alt
4.15/5
MeloTTS
Free
Naturlighed
3.8/5
Nøjagtighed
4.1/5
Hastighed
4.6/5
I alt
4.13/5
Dia TTS
Standard
Naturlighed
4.6/5
Nøjagtighed
4.3/5
Hastighed
3.2/5
I alt
4.09/5
VITS
Free
Naturlighed
3.4/5
Nøjagtighed
4.0/5
Hastighed
4.8/5
I alt
4.0/5
Orpheus
Standard
Naturlighed
4.3/5
Nøjagtighed
4.1/5
Hastighed
3.5/5
I alt
4.0/5
OpenVoice
Premium
Naturlighed
4.0/5
Nøjagtighed
4.1/5
Hastighed
3.9/5
I alt
4.0/5
IndexTTS-2
Standard
Naturlighed
4.3/5
Nøjagtighed
4.1/5
Hastighed
3.2/5
I alt
3.91/5
Spark TTS
Standard
Naturlighed
4.2/5
Nøjagtighed
4.0/5
Hastighed
3.4/5
I alt
3.9/5
Parler TTS
Standard
Naturlighed
4.1/5
Nøjagtighed
3.9/5
Hastighed
3.4/5
I alt
3.83/5
Tortoise TTS
Premium
Naturlighed
4.6/5
Nøjagtighed
4.4/5
Hastighed
1.8/5
I alt
3.7/5
Bark
Standard
Naturlighed
4.2/5
Nøjagtighed
3.8/5
Hastighed
2.5/5
I alt
3.57/5
Benchmarkmetodologi
Testopsætning
- Hardware: 4x NVIDIA Tesla P40 (24GB VRAM hver), 96GB i alt
- Testtekst: 5 standardiserede passager, der dækker forskellige talemønstre (fortælling, dialog, teknisk, emotionel, flersproget)
- Evaluering: Automatiserede målinger (MOS-estimation, WER, RTF) kombineret med lyttetest
- Kører: Hver model testet 10 gange pr. passage, scorer i gennemsnit
Scoringskriterier
- Naturlighed (40%): Prosody, intonation, rytme, følelser. Hvordan lyder det?
- Nøjagtighed (30%): Udtale korrekthed, ordfejlrate, forståelighed
- Hastighed (30%): Real-time faktor (lyd sekunder / generation sekunder). Højere = hurtigere.
- Samlet set: Vejet gennemsnit: 0.4 x Naturlighed + 0.3 x Nøjagtighed + 0.3 x Hastighed
Bemærk: Benchmarks afspejler ydeevnen på vores specifikke hardware og test tekster. Den virkelige verden kvalitet kan variere baseret på input tekst, sprog og stemmevalg. Fællesskab ratings giver et komplementært signal baseret på forskellige reelle brug.
Ofte stillede spørgsmål
Hvad kan vi forbedre? Din feedback hjælper os med at løse problemer.
Find din perfekte stemme
Prøv enhver model gratis med Kokoro, Piper, VITS, eller MeloTTS. Ingen konto kræves.