发表机构
Sestek（Sestek）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究对四个现代TTS系统在土耳其语上的微调和零样本配置进行了系统基准测试，使用十八种客观指标评估，并分析了长度与时间一致性影响，发布了评估代码。
AI 中文摘要
现代文本转语音（TTS）系统可以从短参考片段克隆目标说话人，或针对目标语音进行微调，但它们在土耳其语等形态丰富、资源较少的语言上的行为仍未得到充分刻画。我们提出了一个系统性基准测试，涵盖四个当代系统（Chatterbox、CosyVoice、OmniVoice和VoxCPM2），在微调和零样本配置下进行评估，并与传统的VITS基线进行对比，以自然黄金语音为锚点。每种配置使用十八个互补的客观指标进行评分，涵盖学习型自然度预测器（UTMOS v2、DNSMOS-Pro、SCOREQ、WhisQA、AudioBox-PQ、NatScore、SpeechLMScore）、可懂度和信号质量估计器（SQUIM PESQ/SI-SDR/STOI、Brouhaha）、说话人相似度、分布保真度（TTSDS）和低级声学描述符。我们进一步分析了质量如何随话语长度变化，并通过分块话语的说话人身份和自然度漂移来量化长时时间一致性。我们发布了评估代码，以支持可复现的TTS评估。
英文摘要
Modern text-to-speech (TTS) systems can clone a target speaker from a short reference clip or be fine-tuned on a target voice, yet their behaviour on morphologically rich, lower-resource languages such as Turkish remain under-characterised. We present a systematic benchmark of four contemporary systems (Chatterbox, CosyVoice, OmniVoice, and VoxCPM2) evaluated across fine-tuned and zero-shot configurations, contrasted with a conventional VITS baseline and anchored to natural gold speech. Each configuration is scored with eighteen complementary objective metrics spanning learned naturalness predictors (UTMOS v2, DNSMOS-Pro, SCOREQ, WhisQA, AudioBox-PQ, NatScore, SpeechLMScore), intelligibility and signal-quality estimators (SQUIM PESQ/SI-SDR/STOI, Brouhaha), speaker similarity, distributional fidelity (TTSDS) and low-level acoustic descriptors. We further analyse how quality varies with utterance length and quantify long-form temporal consistency through speaker-identity and naturalness drift over chunked utterances. We release our evaluation code to support reproducible TTS evaluation.
CommentsAccepted at 28th International Conference on Speech and Computer (SPECOM 2026). Published in Lecture Notes in Computer Science
Journal refLecture Notes in Computer Science, SPECOM 2026, Springer, 2026
DOI：10.1007/978-3-032-37867-5_31
