SSR7000 Lightweight Proxies Dataset (v1)

Lightweight video proxies for AVSR experiments derived from the SSR7000 dataset. Contents: - proxies/lip/*.mp4 (HEVC/H.264, 128x128@30fps, grayscale) - proxies/uti/*.mp4 (H.264/H.265, 192x96@30fps, grayscale) - splits/train/text, splits/valid/text (utterance-level transcripts) - COUNTS.json, FILES.txt, sha256sums.txt for integrity and inventory Notes: - `utt` is the unique key aligning videos and transcripts. - Intended for rapid prototyping and benchmarking; full-resolution raw data not included. - See README/CONTENTS.txt inside the archive for layout and verification steps.