Lëscht vu lëtzebuergesche Filmregisseuren Lëscht vun däitsche Filmregisseuren
Verglach méi wéi 20 Text-zu-Sprooch-Modeller. Offiziell Benchmarks, Gemeinschaftswäertungen a Side-by-Side-Vergleich.
Side-by-Side Verglach
Typ Text, wielt zwou Modeller a vergläicht d'Resultater. Free-tier Modeller brauchen kee Kont.
D'Modeller sinn no an no un d'Konkurrenz ausgefall. Anmelden D'Premiere vum Film war den 1.
Lëscht vu Lëtzebuerger Moler
| # | Modell | Offiziell | Gemeinschaft | Är Bewertung | Geschwindegkeet | Tier |
|---|---|---|---|---|---|---|
| 1 |
Kokoro
Lightweight 82M parameter model delivering studio-quality speech with blazing-fast inference.
82M
1200h
2024
|
4.8 /5 |
5.0
/5
1 Stëmm
|
fast | Free | |
| 2 |
CosyVoice 2
Alibaba's scalable streaming TTS with human-parity naturalness and near-zero latency.
300M
200000h
2024
|
4.26 /5 | Noch keng Stëmmen | medium | Standard | |
| 3 |
Chatterbox
State-of-the-art zero-shot voice cloning with emotion control from Resemble AI.
300M
2025
|
4.25 /5 | Noch keng Stëmmen | medium | Premium | |
| 4 |
StyleTTS 2
Human-level text-to-speech through style diffusion and adversarial training.
100M
585h
2024
|
4.23 /5 | Noch keng Stëmmen | medium | Premium | |
| 5 |
Piper
A fast, local neural text to speech system optimized for Raspberry Pi and embedded devices.
15M
2023
|
4.15 /5 | Noch keng Stëmmen | fast | Free | |
| 6 |
MeloTTS
High-quality multilingual text-to-speech that runs on CPU with minimal latency.
25M
2024
|
4.13 /5 | Noch keng Stëmmen | fast | Free | |
| 7 |
Dia TTS
Multi-speaker dialog generation model that creates natural conversations between speakers.
1.6B
2024
|
4.09 /5 | Noch keng Stëmmen | medium | Standard | |
| 8 |
VITS
Conditional variational autoencoder with adversarial learning for end-to-end text-to-speech.
25M
585h
2021
|
4.0 /5 | Noch keng Stëmmen | fast | Free | |
| 9 |
Orpheus
Human-level emotional TTS model trained on 100K hours of speech data.
3B
100000h
2025
|
4.0 /5 | Noch keng Stëmmen | medium | Standard | |
| 10 |
OpenVoice
Instant voice cloning with granular control over style, emotion, and accent.
300M
2024
|
4.0 /5 | Noch keng Stëmmen | medium | Premium | |
| 11 |
IndexTTS-2
Zero-shot TTS with fine-grained emotion control and high expressiveness.
300M
2025
|
3.91 /5 | Noch keng Stëmmen | medium | Standard | |
| 12 |
Spark TTS
Voice cloning TTS with controllable emotion and speaking style via prompts.
500M
2025
|
3.9 /5 | Noch keng Stëmmen | medium | Standard | |
| 13 |
Parler TTS
Describe the voice you want in natural language and Parler generates matching speech.
880M
45000h
2024
|
3.83 /5 | Noch keng Stëmmen | medium | Standard | |
| 14 |
Tortoise TTS
Multi-voice text-to-speech focused on quality with autoregressive architecture.
400M
50000h
2022
|
3.7 /5 | Noch keng Stëmmen | slow | Premium | |
| 15 |
Bark
Transformer-based text-to-audio model that generates realistic speech, music, and sound effects.
350M
100000h
2023
|
3.57 /5 | Noch keng Stëmmen | slow | Standard | |
| 16 |
Bark Small
Lighter version of Bark with faster inference and lower memory usage.
150M
100000h
2023
|
— | Noch keng Stëmmen | medium | Standard | |
| 17 |
GLM-TTS
Achieves the lowest character error rate among open-source TTS models.
300M
2025
|
— | Noch keng Stëmmen | medium | Standard | |
| 18 |
GPT-SoVITS
Few-shot voice cloning TTS that replicates any voice from just 5 seconds of audio.
200M
2024
|
— | Noch keng Stëmmen | slow | Standard | |
| 19 |
Qwen3 TTS
Alibaba's multilingual TTS with voice cloning, preset voices, and voice design from text.
1.7B
2025
|
— | Noch keng Stëmmen | medium | Standard | |
| 20 |
Sesame CSM
Conversational speech model generating natural dialogue with appropriate timing and emotion.
1B
2025
|
— | Noch keng Stëmmen | slow | Premium | |
| 21 |
Chatterbox Turbo
Faster Chatterbox with sub-200ms latency and paralinguistic tags for laughs, coughs, and more.
350M
2025
|
— | Noch keng Stëmmen | fast | Standard | |
| 22 |
Zonos
Emotion-controllable TTS with fine-grained sliders for happiness, anger, sadness, and more.
1.6B
200000h
2025
|
— | Noch keng Stëmmen | medium | Standard | |
| 23 |
Dia 2
Streaming-first conversational TTS with multi-speaker dialogue and paralinguistic cues.
2B
2025
|
— | Noch keng Stëmmen | fast | Standard | |
| 24 |
VoxCPM
Tokenizer-free TTS producing 44.1kHz audio with context-aware paragraph consistency.
500M
1800000h
2025
|
— | Noch keng Stëmmen | fast | Standard | |
| 25 |
OuteTTS
LLM-based TTS that runs on CPU, GPU, or browser via llama.cpp and Transformers.js.
1B
5000h
2025
|
— | Noch keng Stëmmen | fast | Free | |
| 26 |
TADA
Zero-hallucination TTS with text-acoustic dual alignment, 5x faster than comparable LLM TTS.
1B
2026
|
— | Noch keng Stëmmen | fast | Standard | |
| 27 |
VibeVoice
Microsoft's multi-speaker long-form TTS generating up to 90 minutes with 4 distinct speakers.
1.5B
100000h
2025
|
— | Noch keng Stëmmen | fast | Standard | |
| 28 |
Pocket TTS
Lightweight 100M parameter model by Kyutai with voice cloning from a single sample.
100M
50000h
2025
|
— | Noch keng Stëmmen | fast | Free | |
| 29 |
Kitten TTS
Ultra-lightweight TTS under 80MB. Runs on CPU without GPU.
80M
2025
|
— | Noch keng Stëmmen | fast | Free | |
| 30 |
CosyVoice3
Next-generation multilingual TTS with bi-streaming, emotion control, and zero-shot voice cloning.
500M
200000h
2025
|
— | Noch keng Stëmmen | 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
|
— | Noch keng Stëmmen | medium | Premium | |
| 32 |
MegaTTS3
ByteDance's sparse alignment TTS with adjustable intelligibility vs. speaker similarity.
1B
100000h
2025
|
— | Noch keng Stëmmen | slow | Premium |
Lëscht vu lëtzebuergesche Komponisten
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
Lëscht vu Methodologien
Testaufbau
- Hardware: 4x NVIDIA Tesla P40 (24GB VRAM all), 96GB am Ganzen
- Testtext: 5 Standardiséiert Passagen, déi verschidde Sproochmusteren ofdecken (Narrativ, Dialog, technesch, emotional, multilingual)
- Bewäertung: D'Automatiséiert Metriken (MOS Estimatioun, WER, RTF) kombinéiert mat mënschleche Liesversich
- Ausféierungen: All 10 Minutte gëtt en Test gefouert, fir d'Scoren ze evaluéieren.
Kritär vun der Bewäertung
- Natierlechkeet (40%): Emotion, Emotioun, Emotioun, Emotioun, Emotioun, Emotioun, Emotioun, Emotioun, Emotioun, Emotioun.
- Genauegkeet (30%): Aussproochkorrektur, Wuertfehlerquote, Verständlechkeet
- Geschwindegkeet (30%): Echtzäitfaktor (Audiosekonnen / Generéierungssekonnen). Héich = méi séier.
- Gesamt: Duerchschnëttlech Gewiicht: 0,4 x Natierlechkeet + 0,3 x Genauegkeet + 0,3 x Geschwindegkeet
Notiz: Benchmarks reflektéieren d'Performance op eiser spezifescher Hardware an Testtexter. D'Realweltqualitéit kann ofhängeg vum Inputtext, der Sprooch an der Stëmmauswiel variéieren. D'Community-Bewäertungen bidden e komplementäre Signal baséiert op verschiddener realer Benotzung.
Häufig gestallte Froen
What could we improve? Your feedback helps us fix issues.
Den Haaptuert ass Perpignan.
Probéiert all Modell gratis mat Kokoro, Piper, VITS oder MeloTTS. Keng Kont néideg.