Nemotron 3 Diarization (MLX)

Converted from nvidia/Nemotron-3-Diarization for use with mlx-audio.

Use the mlx_audio package (MLX-Audio). The MLX checkpoint is copied from mlx-community/Nemotron-3-Diarization.

It predicts up to eight speakers at 10 ms resolution from 16 kHz mono audio.

from mlx_audio.vad import load

model = load("nativ-community/Nemotron-3-Diarization", strict=True)
result = model.generate("meeting.wav")
print(result.text)

# Incremental results retain speaker identities through the AOSC and FIFO.
for result in model.generate_stream("meeting.wav"):
    for segment in result.segments:
        print(segment.start, segment.end, segment.speaker)

For live PCM input, call model.feed(chunk, state) with a state from model.init_streaming_state() and 16 kHz mono chunks. Flush the final partial chunk and lookahead with model.feed([], state, final=True). Timestamps are absolute within the recording. Labels are generic arrival-order speaker IDs; the model does not identify people. Overlapping speakers may be active together.

Consult the upstream model card for training data, evaluation and license information.

Downloads last month: 32

Safetensors

Model size

99.3M params

Tensor type

BF16

F32

MLX

Hardware compatibility

Quantized

Inference Providers NEW

Voice Activity Detection

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for nativ-community/Nemotron-3-Diarization

Base model

nvidia/Nemotron-3-Diarization

Finetuned

(15)

this model