taejinp's picture
Update Nemotron streaming multispeaker ASR Space
2dd1cd7 verified
Raw
History Blame Contribute Delete
2.69 kB
<div style='text-align: right;'>
<h1 style='text-align: right; color: #76B900; margin: 0;'>Multitalker 8-spk Streaming Transcription</h1>
<p>
Overlap-aware streaming ASR with <a href='https://huggingface.co/nvidia/multitalker-parakeet-streaming-0.6b-v1' target='_blank'>Multitalker Parakeet</a> and <a href='https://huggingface.co/nvidia/Nemotron-3-Diarization-preview' target='_blank'>Nemotron Diarization</a>
</p>
<!-- <p><strong style='color: #76B900; font-size: 1.5em;'>Parameters</strong></p> -->
<!-- <table style='font-size: 1.1em; display: inline-block; text-align: left; border-collapse: collapse; width: 98%; border: 1px solid #3d3d3d;'>
<tr>
<td style='padding: 10px; vertical-align: top; width: 50%; border: 1px solid #3d3d3d;'>
<strong>Audio Pre-loading:</strong> If enabled, indicates the time when audio is fully loaded.
</td>
<td style='padding: 10px; vertical-align: top; width: 50%; border: 1px solid #3d3d3d;'>
<strong>Streaming Chunk Size:</strong> Represents a trade-off between latency and accuracy.
</td>
</tr>
<tr>
<td style='padding: 10px; vertical-align: top; border: 1px solid #3d3d3d;'>
<strong>Number of Speakers:</strong> The system ignores any speakers exceeding this limit.
</td>
<td style='padding: 10px; vertical-align: top; border: 1px solid #3d3d3d;'>
<strong>Single Speaker Mode:</strong> If enabled, functions as a single-speaker ASR model.
</td>
</tr>
<tr>
<td style='padding: 10px; vertical-align: top; border: 1px solid #3d3d3d;'>
<strong>Cache Gating:</strong> Stops updating the cache if no speech activity detected.
</td>
<td style='padding: 10px; vertical-align: top; border: 1px solid #3d3d3d;'>
<strong>Binary Diarization Decision:</strong> If enabled, binarizes the diarization logits.
</td>
</tr>
<tr>
<td style='padding: 10px; vertical-align: top; border: 1px solid #3d3d3d;'>
<strong>Text Break Silence:</strong> Inserts line breaks for silences longer than this value.
</td>
<td style='padding: 10px; vertical-align: top; border: 1px solid #3d3d3d;'>
<strong>Displayed Transcript Lines:</strong> Number of transcript lines displayed.
</td>
</tr>
</table> -->
<!--
<p>
<a href='https://huggingface.co/nvidia/multitalker-parakeet-streaming-0.6b-v1' target='_blank'>πŸ€— Hugging Face Model Page</a> |
<a href='https://arxiv.org/abs/2506.22646' target='_blank'>πŸ“„ Paper</a> |
<a href='https://github.com/NVIDIA/NeMo' target='_blank'>πŸš€ NeMo Repository</a>
</p>
-->
</div>