baari 0.1.0

End-of-turn detection for Urdu voice agents.
At every pause, it decides whether the speaker has finished (reply now) or is only pausing mid-sentence (keep listening).

baari (باری) means "turn" in Urdu. The table below shows that today's turn detectors often cut Urdu speakers off mid-sentence, or wait after they have finished. baari is a small audio model trained to get this right. It follows the Smart Turn recipe and takes the same input and output, so it can stand in for Smart Turn.

Status: work in progress. Done so far: the Urdu test set, the scores of existing models, and labels for 61,251 training segments. The training code is in place. The first model hasn't been trained yet.

🎧 Do existing models understand Urdu?

1,588 clips of real Urdu phone conversations, cut from the official held-out split of IndicVoices: 1,470 where the speaker had finished and 118 where they were pausing mid-sentence. Because most clips are "finished", a model that always says "finished" gets 92.6% plain accuracy, so the fair score is the average of the two middle columns (balanced accuracy).

Scored on a MacBook Pro (M3 Pro), with the same audio preparation Pipecat uses. int8 models are sensitive: their scores shift by a few points between Arm and x86 CPUs, and even tiny changes in how the audio is prepared flip about 1 answer in 100. fp32 results stay the same.

Use it

pip install baari

from baari import TurnDetector

detector = TurnDetector(threads=4)  # downloads baari's 9 MB int8 model once
detector.is_finished(audio)  # 16 kHz mono audio (float or int16) that ends at the pause
detector.probability(audio)  # the same decision as a number from 0 to 1

It needs only numpy and onnxruntime. One call takes about 12 ms on 4 CPU threads and 37 ms on one (MacBook Pro, M3 Pro). Use threads=1 on a server that handles many calls at once.

The baari model is not released yet. Until it is, TurnDetector(path=...) runs any Smart Turn style ONNX file, such as Indic Smart Turn tiny, the best existing model in the table above.

🛠️ Run it

uv sync --extra pipeline
echo "HF_TOKEN=..." > .env          # accept the IndicVoices terms on Hugging Face first
uv run baari scan --split valid     # index the conversation segments (text only)
uv run baari batches                # label batches for the LLM agents
# the agents write data/labels/valid/results/{a,b}/batch-*.jsonl
uv run baari build                  # cut the 8 s clips
uv run baari evaluate               # score the existing models

uv run baari scan --split train
uv run baari batches --split train
# the agents write data/labels/train/results/a/batch-*.jsonl
uv run baari build --split train --labelers a --delete-shards
uv run baari train                  # fine-tune Indic Smart Turn tiny, export fp32 and int8 ONNX

Credits · Contributing · BSD-2-Clause license
Made by Abdul Samad (abdulsamadsid1@gmail.com)