Romanian TTS listening grid

Fifteen systems, seventeen sentences, 255 clips — the complete publishable set from the survey of 2026-08-18, ordered by blind listening score. Systems marked cloned were given the same Romanian reference clip; own voice systems cannot clone and speak their own; description and voice design systems take a written description of a voice instead of audio.

All audio here is synthetic — text-to-speech output, not a recording of anyone. Published for research and comparative evaluation, non-commercially. See LICENSES.md.

a01_salut everyday

Bună dimineața! Afară sunt paisprezece grade și e înnorat.

the sentence the blind scores were given on

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 4.0s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 4.2s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 4.7s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 3.6s
Piper Liana high 9/10 · own voice · 22.05 kHz · 3.8s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 3.9s
Supertonic 3 9/10 · own voice · 44.1 kHz · 5.8s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 4.6s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 3.8s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 3.5s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 4.4s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 5.2s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 4.1s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 5.0s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 5.4s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 4.8s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 4.8s
Maya1 2/10 · description · 24 kHz · 6.0s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 4.0s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 4.8s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 3.8s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 3.7s
Maya1 — English control 1/10 · description · 24 kHz · 4.3s

a02_casa everyday

Am stins luminile din living și am încuiat ușa de la intrare.

a second neutral sentence

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 3.7s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 4.1s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 3.8s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 3.7s
Piper Liana high 9/10 · own voice · 22.05 kHz · 3.8s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 3.5s
Supertonic 3 9/10 · own voice · 44.1 kHz · 5.2s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 4.1s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 3.5s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 3.6s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 3.8s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 4.7s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 3.7s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 5.3s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 4.4s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 4.0s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 5.8s
Maya1 2/10 · description · 24 kHz · 5.6s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 5.1s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 5.4s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 5.1s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 4.0s
Maya1 — English control 1/10 · description · 24 kHz · 3.2s

a03_numerale everyday

Am pornit centrala. Ajunge la douăzeci și unu de grade în vreo douăzeci de minute.

numerals written as words — text normalisation shows up here

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 5.1s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 5.0s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 5.0s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 5.0s
Piper Liana high 9/10 · own voice · 22.05 kHz · 4.5s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 4.5s
Supertonic 3 9/10 · own voice · 44.1 kHz · 6.8s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 4.8s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 4.8s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 4.8s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 4.7s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 6.0s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 5.2s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 6.4s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 5.4s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 6.4s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 7.0s
Maya1 2/10 · description · 24 kHz · 7.3s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 6.4s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 6.6s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 5.1s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 4.8s
Maya1 — English control 1/10 · description · 24 kHz · 5.7s

a04_sedinta everyday

Ședința este programată la ora trei, marți după-amiază.

opens on a capital Ș

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 3.4s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 3.7s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 4.0s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 3.5s
Piper Liana high 9/10 · own voice · 22.05 kHz · 4.0s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 3.9s
Supertonic 3 9/10 · own voice · 44.1 kHz · 5.0s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 4.0s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 3.4s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 3.4s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 3.6s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 4.1s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 3.6s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 4.3s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 4.2s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 4.3s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 4.7s
Maya1 2/10 · description · 24 kHz · 4.9s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 4.6s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 4.3s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 4.2s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 4.5s
Maya1 — English control 1/10 · description · 24 kHz · 3.4s

a05_diacritice everyday

Documentul are șaptesprezece pagini și trebuie citit până joi.

diacritic-dense: ș, ț, ă, î

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 3.8s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 4.1s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 4.1s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 3.9s
Piper Liana high 9/10 · own voice · 22.05 kHz · 3.9s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 3.9s
Supertonic 3 9/10 · own voice · 44.1 kHz · 5.2s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 3.9s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 4.2s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 3.7s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 4.2s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 5.6s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 4.0s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 4.9s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 9.4s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 4.5s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 6.7s
Maya1 2/10 · description · 24 kHz · 4.9s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 6.0s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 5.0s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 4.0s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 4.1s
Maya1 — English control 1/10 · description · 24 kHz · 5.2s

a06_loanwords everyday

Am primit un e-mail de la Google despre update-ul de software.

English loanwords in Romanian morphology

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 3.4s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 3.8s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 3.7s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 3.8s
Piper Liana high 9/10 · own voice · 22.05 kHz · 3.2s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 3.4s
Supertonic 3 9/10 · own voice · 44.1 kHz · 5.1s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 3.6s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 3.5s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 3.5s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 3.0s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 3.9s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 3.2s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 4.6s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 5.4s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 3.5s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 5.0s
Maya1 2/10 · description · 24 kHz · 4.5s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 5.2s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 4.1s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 3.5s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 4.8s
Maya1 — English control 1/10 · description · 24 kHz · 3.4s

p01_punct_punct punctuation

Te rog, spune-mi că nu vorbești serios.

the same clause, four terminal marks — full stop

Marked : this system rendered byte-identical audio for two or more of the four terminal marks — its text front end destroyed the distinction before the model saw it. Here: FireRedTTS3, MMS-TTS Romanian, FireRedTTS3-Instruct.

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 2.4s ≡ 2
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 2.7s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 2.6s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 2.4s
Piper Liana high 9/10 · own voice · 22.05 kHz · 2.6s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 2.5s
Supertonic 3 9/10 · own voice · 44.1 kHz · 3.3s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 2.0s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 2.7s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 2.3s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 2.2s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 3.2s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 2.7s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 3.0s ≡ 4
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 3.3s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 2.9s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 3.6s
Maya1 2/10 · description · 24 kHz · 4.3s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 3.5s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 3.4s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 3.7s ≡ 2
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 3.3s
Maya1 — English control 1/10 · description · 24 kHz · 3.2s

p02_punct_intreb punctuation

Te rog, spune-mi că nu vorbești serios?

question mark

Marked : this system rendered byte-identical audio for two or more of the four terminal marks — its text front end destroyed the distinction before the model saw it. Here: MMS-TTS Romanian.

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 2.4s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 2.6s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 2.6s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 2.4s
Piper Liana high 9/10 · own voice · 22.05 kHz · 2.2s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 2.4s
Supertonic 3 9/10 · own voice · 44.1 kHz · 3.3s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 2.9s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 5.1s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 2.3s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 2.6s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 2.8s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 2.6s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 3.0s ≡ 4
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 3.3s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 2.7s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 4.5s
Maya1 2/10 · description · 24 kHz · 4.2s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 3.2s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 3.4s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 3.5s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 3.5s
Maya1 — English control 1/10 · description · 24 kHz · 3.2s

p03_punct_excl punctuation

Te rog, spune-mi că nu vorbești serios!

exclamation mark

Marked : this system rendered byte-identical audio for two or more of the four terminal marks — its text front end destroyed the distinction before the model saw it. Here: MMS-TTS Romanian.

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 2.2s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 2.5s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 2.4s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 2.4s
Piper Liana high 9/10 · own voice · 22.05 kHz · 2.2s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 2.5s
Supertonic 3 9/10 · own voice · 44.1 kHz · 3.5s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 2.5s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 5.8s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 2.3s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 2.2s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 3.1s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 2.6s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 3.0s ≡ 4
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 2.8s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 2.9s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 4.9s
Maya1 2/10 · description · 24 kHz · 4.6s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 3.4s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 3.4s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 3.7s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 3.3s
Maya1 — English control 1/10 · description · 24 kHz · 3.3s

p04_punct_puncte punctuation

Te rog, spune-mi că nu vorbești serios...

ellipsis

Marked : this system rendered byte-identical audio for two or more of the four terminal marks — its text front end destroyed the distinction before the model saw it. Here: FireRedTTS3, MMS-TTS Romanian, FireRedTTS3-Instruct.

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 2.4s ≡ 2
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 2.6s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 2.5s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 2.5s
Piper Liana high 9/10 · own voice · 22.05 kHz · 2.0s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 2.3s
Supertonic 3 9/10 · own voice · 44.1 kHz · 3.9s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 2.5s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 9.1s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 2.4s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 2.2s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 2.3s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 2.6s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 3.0s ≡ 4
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 12.4s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 3.4s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 3.9s
Maya1 2/10 · description · 24 kHz · 3.8s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 4.0s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 3.4s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 3.7s ≡ 2
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 3.7s
Maya1 — English control 1/10 · description · 24 kHz · 3.1s

p05_intrebare punctuation

Nu am înțeles. Poți să repeți, te rog?

a two-clause question

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 2.7s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 2.7s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 2.8s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 2.3s
Piper Liana high 9/10 · own voice · 22.05 kHz · 2.5s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 2.7s
Supertonic 3 9/10 · own voice · 44.1 kHz · 3.7s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 3.4s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 2.9s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 2.4s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 2.9s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 3.5s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 2.7s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 3.6s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 4.2s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 3.8s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 3.6s
Maya1 2/10 · description · 24 kHz · 3.6s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 3.6s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 3.8s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 3.5s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 3.0s
Maya1 — English control 1/10 · description · 24 kHz · 3.9s

e01_emotiv expressive

Nu îmi vine să cred că s-a întâmplat asta.

emotive content with no emotion markup

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 2.2s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 2.5s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 2.1s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 2.6s
Piper Liana high 9/10 · own voice · 22.05 kHz · 2.0s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 2.1s
Supertonic 3 9/10 · own voice · 44.1 kHz · 3.6s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 2.0s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 2.6s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 2.6s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 2.4s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 8.7s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 2.4s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 3.0s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 3.0s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 3.2s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 4.9s
Maya1 2/10 · description · 24 kHz · 4.5s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 4.2s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 5.0s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 3.0s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 3.1s
Maya1 — English control 1/10 · description · 24 kHz · 2.6s

s01_vulpea story

— Bun! gândi vulpea. Iaca hrana ce-o așteptam eu.

literary Romanian, opens on a dialogue dash

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 3.5s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 4.2s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 4.7s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 2.9s
Piper Liana high 9/10 · own voice · 22.05 kHz · 3.5s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 3.9s
Supertonic 3 9/10 · own voice · 44.1 kHz · 5.2s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 4.1s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 3.7s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 3.0s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 3.0s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 4.5s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 3.4s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 3.9s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 4.1s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 4.5s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 5.5s
Maya1 2/10 · description · 24 kHz · 8.8s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 4.0s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 4.8s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 5.0s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 3.4s
Maya1 — English control 1/10 · description · 24 kHz · 5.7s

s02_taranul story

— Bre! da' cum naiba a murit vulpea asta aici?! Ti!... ce frumoasă cațaveică am să fac nevestei mele din blana istui vulpoiu!

regionalisms and stacked punctuation

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 7.4s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 8.3s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 10.5s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 7.2s
Piper Liana high 9/10 · own voice · 22.05 kHz · 7.3s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 7.5s
Supertonic 3 9/10 · own voice · 44.1 kHz · 12.4s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 8.1s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 9.4s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 7.2s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 6.9s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 10.3s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 7.2s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 10.3s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 8.6s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 8.6s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 9.0s
Maya1 2/10 · description · 24 kHz · 13.9s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 9.0s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 9.2s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 10.2s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 5.5s
Maya1 — English control 1/10 · description · 24 kHz · 9.4s

s03_cucurigu story

— Cucurigu! boieri mari, dați punguța cu doi bani!

short exclamatory line

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 3.4s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 3.4s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 3.8s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 3.0s
Piper Liana high 9/10 · own voice · 22.05 kHz · 3.4s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 3.5s
Supertonic 3 9/10 · own voice · 44.1 kHz · 5.3s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 4.3s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 3.8s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 3.0s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 3.4s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 5.1s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 3.6s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 4.4s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 3.5s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 4.0s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 4.8s
Maya1 2/10 · description · 24 kHz · 7.8s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 5.5s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 4.1s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 3.7s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 3.0s
Maya1 — English control 1/10 · description · 24 kHz · 7.4s

s04_baba story

— Măi babă, mănânci ca în târgul lui Cremene. Ia dă-mi și mie niște ouă, ca să-mi prind pofta măcar.

longer literary passage

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 5.6s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 6.0s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 6.3s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 5.7s
Piper Liana high 9/10 · own voice · 22.05 kHz · 6.3s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 5.9s
Supertonic 3 9/10 · own voice · 44.1 kHz · 8.1s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 6.6s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 5.9s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 6.3s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 5.8s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 6.9s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 6.7s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 8.3s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 6.0s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 7.4s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 8.4s
Maya1 2/10 · description · 24 kHz · 9.9s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 7.2s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 8.6s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 7.4s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 4.9s
Maya1 — English control 1/10 · description · 24 kHz · 11.3s

L01_lung long

Era odată o vulpe vicleană, ca toate vulpile. Ea umblase o noapte întreagă după hrană și nu găsise nicăiri. […] Carul apropiindu-se de vulpe, țăranul ce mâna boii o vede și, crezând că-i moartă cu adevărat, strigă la boi. Boii se opresc.

559 characters — exposes drift and truncation

FireRedTTS3 10/10 · cloned from Liana · 24 kHz · 35.3s
Higgs TTS 3 9/10 · cloned from Liana · 24 kHz · 38.7s
NeuTTS Romanian 9/10 · cloned from Liana · 24 kHz · 40.0s
OmniVoice 9/10 · cloned from Liana · 24 kHz · 32.6s
Piper Liana high 9/10 · own voice · 22.05 kHz · 35.9s
Piper Liana medium 9/10 · own voice · 22.05 kHz · 37.0s
Supertonic 3 9/10 · own voice · 44.1 kHz · 43.4s
ZONOS2 9/10 · cloned from Kore · 44.1 kHz · 43.3s
dots.tts-soar 8/10 · cloned from Liana · 48 kHz · 29.9s
F5-TTS Romanian 7/10 · cloned from Liana · 24 kHz · 32.4s
MOSS-TTS-Local v1.5 7/10 · cloned from Liana · 48 kHz · 37.4s
VibeVoice-1.5B 7/10 · cloned from Kore · 24 kHz · 38.7s
Fish S2 Pro 6/10 · cloned from Liana · 44.1 kHz · 34.7s
MMS-TTS Romanian 5/10 · own voice · 16 kHz · 42.7s
TADA-3B-ML 4/10 · cloned from Kore · 24 kHz · 42.2s
VoxCPM2 4/10 · cloned from Kore · 48 kHz · 22.1s
Chatterbox 2/10 · cloned from Kore · 24 kHz · 32.4s
Maya1 2/10 · description · 24 kHz · 36.4s
CosyVoice3 0.5B 1/10 · cloned from Kore · 24 kHz · 32.3s
Dia2-2B 1/10 · cloned from Kore · 24 kHz · 42.6s
FireRedTTS3-Instruct 1/10 · voice design · 24 kHz · 38.0s
IndexTTS-2.5 1/10 · cloned from Kore · 22.05 kHz · 31.7s
Maya1 — English control 1/10 · description · 24 kHz · 36.4s