Curated Multimodal Datasets & safetensors
Open training corpora across high-res video tokens, phoneme sequences, and FP8 model weights.
| Dataset Key | Modality | Tokens / Records | Format | Size | Download Shard |
|---|---|---|---|---|---|
| synapse-open-video-10m | 4K Video Latents | 10.4M Clips (60fps) | WebDataset HDF5 | 2.4 TB | Fetch Shard #01 (10 MB) |
| synapse-acoustic-phoneme-v3 | Multilingual Speech | 850,000 Hours | Parquet / PCM | 480 GB | Fetch Shard #01 (10 MB) |