DNU-ViSpeech is a Vietnamese scripted/read-speech corpus for automatic speech recognition (ASR). This release contains 31,070 utterances from 26 anonymized speakers, totaling 75.570 hours. Audio is WAV, 16 kHz, mono, uncompressed PCM 16-bit. The primary protocol is DNU-ASR-v1.1. Its evaluation split combines unseen-speaker and recording-configuration shift; those two factors cannot be isolated with the current data. The corpus supports research on Vietnamese ASR adaptation, especially whether compact models can approach larger zero-shot systems after domain fine-tuning on read speech. The repository follows the Hugging Face AudioFolder layout. Each split has its own metadata.csv, and…