openresearchtools/diar_streaming_sortformer_4spk-v2

Real-time local speaker diarization on a GGUF runtime path

Unofficial community GGUF conversion for real-time local diarization

Derived from nvidia/diar_streaming_sortformer_4spk-v2 and integrated with Openresearchtools-Engine.

---

Item	Details
Upstream model	nvidia/diar_streaming_sortformer_4spk-v2
Conversion source	Derived from the upstream model and converted to GGUF for local runtime use
Conversion script	convert_nemo_sortformer_to_gguf.py
License	CC BY 4.0
How to try it
Desktop app example	Try Transcribe Offline to run live (streaming) or static (from audio files) transcriptions with optional diarization on Mac or Windows (Supports Cuda, Vulkan, Metal graphics acceleration or CPU runtime).
Embeddable runtime	Use Openresearchtools-Engine to integrate real-time local diarization into your own application.

---

Overview

This repository provides an unofficial GGUF conversion derived from nvidia/diar_streaming_sortformer_4spk-v2, converted using convert_nemo_sortformer_to_gguf.py for real-time local speaker diarization with Openresearchtools-Engine.

The original model is distributed under the CC BY 4.0, and these converted weights are distributed subject to that same license.

---

Disclaimer

This is a community conversion and example implementation maintained by OpenResearchTools.

It is not affiliated with, endorsed by, or released by the original model authors or upstream repository owners.

---

Built for real-time local audio pipelines

• Streaming / session-based speaker diarization
• Low-latency local speaker segmentation
• Embeddable runtime integration for diarization-enabled apps

---

Related

Need transcription too?
Pair this with openresearchtools/Voxtral-Mini-4B-Realtime-2602.

Want to test the full OpenResearchTools implementation?
Try Transcribe Offline.

Want to embed the OpenResearchTools runtime?
Use Openresearchtools-Engine.