| license: mit | |
| tags: | |
| - audio | |
| - speech | |
| - neural-audio-codec | |
| - variable-frame-rate | |
| - audio-tokenization | |
| # DTM-Codec | |
| Checkpoint artifacts for [DTM-Codec: Dynamic Token Masking for VFR Speech Coding with Efficient Boundary Selection](https://arxiv.org/abs/2606.29480). | |
| Code and usage instructions are available at [hoyso48/DTM-Codec](https://github.com/hoyso48/DTM-Codec). | |
| | Token rate | Revision | | |
| |---:|---| | |
| | 25 Hz | `25hz` | | |
| | 40 Hz | `40hz` | | |
| | 50 Hz | `50hz` | | |
| | 80 Hz | `80hz` | | |
| The `main` revision mirrors `50hz`. Each revision contains `model.safetensors`, `config.json`, and `conversion.json`. | |