diar streaming sortformer 4spk v2.1 needs about 0.8 GB of memory at F16 (a 0.2 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 0.1 GB (Q8_0) to 0.5 GB (F32).
Streaming speaker diarizer for up to 4 speakers
diar streaming sortformer 4spk v2.1 is a streaming speaker diarization model based on Sortformer architecture. It outputs speaker segments for 16 kHz mono WAV files and does not produce text. The model is trained with an Arrival-Order Speaker Cache for online operation.
Good at
- Online diarization of up to 4 speakers
- FastConformer encoder with 18‑layer Transformer head
- Per‑frame speaker‑activity probabilities
Limits
- No transcription capability
- Requires 16 kHz mono WAV input
- License is NVIDIA Open Model License
Best for
- Real‑time meeting diarization
- Voice agent speaker recognition
- Audio analytics pipelines
Who it’s for. Developers building real‑time voice applications that need speaker identification without transcription. It is suitable for integration with NVIDIA NeMo or transcribe.cpp workflows.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q8_0 | 0.1 GB | 0.7 GB | Fits | Get file → | |
| F16 | 0.2 GB | 0.8 GB | Fits | Get file → | |
| F32 | 0.5 GB | 1.0 GB | Fits | Get file → |
For 8 GB VRAM use Q8_0 (file size 0.14 GB, estimated memory 0.7 GB at 8k context). For 4 GB VRAM use F16 (0.24 GB, 0.8 GB memory). F32 requires 1 GB memory and 0.47 GB file size.
What GPU do you need for diar streaming sortformer 4spk v2.1?
These figures add the file size, an fp16 KV cache at an 8K context window and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is F32 (0.5 GB file, about 1.0 GB in use).
Which diar streaming sortformer 4spk v2.1 quant should you choose?
F16 (0.2 GB) is the middle option in this repository. The unquantized model. It gives original quality and needs the most memory, so it is mainly for conversion or reference.
If memory is tight, the smallest file is Q8_0 at 0.1 GB, with the quality trade-offs that come with fewer bits. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Frequently asked questions about diar streaming sortformer 4spk v2.1
How much VRAM does diar streaming sortformer 4spk v2.1 need?
About 0.8 GB at F16 with an 8K context window. That is the 0.2 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run diar streaming sortformer 4spk v2.1 on an 8 GB GPU?
Yes, with F32: the 0.5 GB file needs about 1.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of diar streaming sortformer 4spk v2.1 GGUF?
It depends on the quant: Q8_0 is 0.1 GB, F16 is 0.2 GB, F32 is 0.5 GB.
Where can I download diar streaming sortformer 4spk v2.1 GGUF?
The files are on Hugging Face in the handy-computer/diar_streaming_sortformer_4spk-v2.1-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does diar streaming sortformer 4spk v2.1 use?
Hugging Face lists the license as other. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Q8_0 on 8 GB VRAM?
Yes, Q8_0 uses about 0.7 GB memory at 8k context, well within 8 GB VRAM.
Does this model produce text?
No, it outputs only speaker segments with start, end, and speaker ID.
What audio format is required?
The model expects 16 kHz mono WAV files; convert other formats before use.