发表机构
BitmanagerAI; lab260; MTUCI（BitmanagerAI; lab260; 莫斯科通信与信息技术大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出俄语长篇语音语料库Balalaika-Longform，通过长单元微调显著降低长文本合成中的词错误率，验证了训练序列长度对连续生成质量的关键作用。
AI 中文摘要
长篇文本转语音必须在长时间生成过程中保留所要求的词语，然而基于句子的训练和评估可能会掩盖词语遗漏和提前停止的问题。我们推出了Balalaika-Longform，这是一个开放的俄语语料库，包含189小时的连续语音单元，时长从30秒到15分钟不等。长单元和匹配的短窗口支持在同一源录音上进行微调比较，随附的评估保留了每一次合成尝试。我们在两种视图上分别对CosyVoice3、Qwen3-TTS、VoxCPM2和F5-TTS进行微调，并在50对文本-语音对上进行连续合成测试，这些语音在微调中未见，测试长度约为75、300和1200个词。在1200个词时，CosyVoice3的词错误率（WER）从短窗口微调后的99.9%降至长目标下的47.3%，在使用带标点和格式匹配的转录文本时进一步降至16.6%；配对区间显示，Qwen3-TTS和VoxCPM2更偏好长目标，F5-TTS也略微偏好长目标，但其绝对WER仍高于90%。这些结果在固定的连续生成协议下，隔离了训练序列长度的影响。
英文摘要
Long-form text-to-speech must retain requested words over extended generations, yet sentence-level training and evaluation can hide omissions and early stops. We introduce Balalaika-Longform, an open Russian corpus of 189 hours in continuous units of 30 seconds to 15 minutes. Long units and matched short windows support fine-tuning comparisons on the same source recordings, and the accompanying evaluation retains every synthesis attempt. We fine-tune CosyVoice3, Qwen3-TTS, VoxCPM2 and F5-TTS on either view and test continuous synthesis on 50 text-voice pairs with voices unseen in fine-tuning, at about 75, 300 and 1,200 words. At 1,200 words, CosyVoice3 WER falls from 99.9% after short-window fine-tuning to 47.3% with long targets, and to 16.6% with punctuated, format-matched transcripts; paired intervals favor long targets for Qwen3-TTS and VoxCPM2, and by a small margin for F5-TTS, whose absolute WER stays above 90%. The results isolate the effect of training sequence length under a fixed continuous-generation protocol.
CommentsSubmitted to IEEE ICASSP 2027. Dataset: https://huggingface.co/datasets/lab260/Balalaika-longform ; code: https://github.com/lab260ru/balalaika-longform
