| --- |
| license: other |
| license_name: nvidia-open-model-license |
| license_link: https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/ |
| base_model: nvidia/diar_streaming_sortformer_4spk-v2.1 |
| tags: |
| - speaker-diarization |
| - sortformer |
| - onnx |
| - streaming |
| --- |
| |
| # Ultra Sortformer 8-speaker (streaming) — ONNX export for TalkBar |
|
|
| ONNX export of the Ultra Sortformer 8-speaker streaming diarization model, |
| used by the TalkBar desktop app as its optional large-meeting (5–8 speaker) |
| live diarizer. |
|
|
| ## Provenance |
|
|
| | Stage | Artifact | License | |
| | --- | --- | --- | |
| | Base model | [nvidia/diar_streaming_sortformer_4spk-v2.1](https://huggingface.co/nvidia/diar_streaming_sortformer_4spk-v2.1) | [NVIDIA Open Model License](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/) | |
| | 8-speaker derivative | [mago-ai/ultra_diar_streaming_sortformer_8spk_v1](https://huggingface.co/mago-ai/ultra_diar_streaming_sortformer_8spk_v1) (`.nemo`) | labeled Apache-2.0 by its authors; the NVIDIA base terms flow through to derivatives | |
| | This repo | `diar_streaming_sortformer_8spk-v1.onnx` — format conversion (`.nemo` → ONNX) of the above, no weight changes beyond export | distributed under the NVIDIA Open Model License terms inherited from the base model | |
|
|
| TalkBar makes **no independent license claim** over these weights: they are |
| a derivative of NVIDIA's streaming Sortformer and are redistributed under |
| the NVIDIA Open Model License, with attribution to NVIDIA and to the |
| Ultra-Sortformer project. Consult the license link above before reuse. |
|
|
| ## File |
|
|
| | File | Bytes | SHA-256 | |
| | --- | ---: | --- | |
| | `diar_streaming_sortformer_8spk-v1.onnx` | 492,246,192 | `f8dda3268c0a7009446b44edea789b25e4a01aae921c28da712050c74b75a252` | |
|
|
| The TalkBar app downloads this file at a pinned commit and verifies the |
| SHA-256 before installing. |
|
|
