# Shruti Bengali

Canonical: https://udgamlabs.com/models/shruti/bengali · Updated: 2026-11-03

Models · Shruti · 29.7 MB · released 3 November 2026

Speech to text for Bengali and English, on your device, with no server.

Try it in your browser: [https://udgamlabs.com/models/shruti/bengali#try](https://udgamlabs.com/models/shruti/bengali#try)

Shruti Bengali · 29.7 MB · Bengali and English · 3 November 2026

## On conversations that mix in English, fewer errors than Bengali models three and five times its size.

Shruti Bengali writes Bengali in Bengali script and keeps English words in English, the way people type: বাচ্চাদের রবীন্দ্রসঙ্গীতের rehearsal শনিবার. On real conversations that mix in English, scored with English words written in English, it makes fewer errors than Vosk Bengali (90 MB) and IndicConformer Bengali (138 MB): DiarBench 36.0 against 42.8 and 39.3. It writes about half of the English words in English letters; both of those models write none. It runs offline on a five-year-old phone, 12.6× faster than real time in 64 MB of memory with the native engine, and it invented no words on 100 silent clips.

| Model | File | DiarBench | FLEURS | Kathbath |
| --- | --- | --- | --- | --- |
| Shruti Bengali | 29.7 MB | 36.0 | 29.9 | 22.2 |
| Whisper tiny | 43.5 MB | 540.9 | 361.6 | 403.7 |
| Whisper base | 81.8 MB | 339.9 | 176.7 | 214.5 |
| Vosk Bengali int8 (third-party copy) | 28.6 MB | 43.3 | 20.9 | 22.4 |
| Vosk Bengali (Zipformer) | 90.0 MB | 42.8 | 20.8 | 21.7 |
| Dolphin-base int8 | 103.7 MB | 93.0 | 41.1 | 45.9 |
| IndicConformer Bengali | 137.7 MB | 39.3 | 20.7 | 12.1 |

Word error rate %, lower is better; 200 clips per set, the same scorer for every model, measured 2 November 2026. DiarBench is conversation that mixes in English, scored with English words written in English; FLEURS and Kathbath are read Bengali. With the English words dropped from the DiarBench reference, both larger models are ahead (34.8 and 38.5 against 41.8). SraVaani-1.0 (909 MB) is ahead on every set (DiarBench 34.6). Whisper adds words, so its rates pass 100%.

### Where it falls short

- On read Bengali, Vosk Bengali and IndicConformer Bengali are clearly more accurate (FLEURS 20.8 and 20.7 against 29.9), and so is a 28.6 MB int8 copy of Vosk Bengali (20.9), slightly smaller than Shruti Bengali.
- Some English words come out in Bengali script (derby as ডার্ভিটা), and spelling variants count as errors.
- Noise, music and speaker-played audio were not tested.

[Try Shruti Bengali](#try) [Shruti Bengali on Hugging Face](https://huggingface.co/udgamlabs/shruti-bengali) the model file, the full results and how each model was run

## Use it in your app

Every Shruti model runs in Cactus’s open-source needle engine. Download the engine for your platform and the file for your language, and pass it a WAV file or 16 kHz audio.

```
# Command line (Mac, Linux, Windows)
# Engine: huggingface.co/Cactus-Compute/needle3, your platform
# Model:  shruti-bengali.cact, from Shruti Bengali's Hugging Face repo
./needle --model shruti-bengali.cact --audio clip.wav
# for speech that is only English: --audio-language en

# Python
pip install cactus-needle

import needle
shruti = needle.Whistle(weights="shruti-bengali.cact")
print(shruti.transcribe("clip.wav")["text"])
```

- Android, iOS and the web: call the engine’s C API. `needle_load` takes the model file, `needle_transcribe` a clip, and `needle_stream_transcribe_process` about a second of live audio at a time. The demo on this page makes the same calls, from WebAssembly in a Web Worker.
- Leave the language unset for Hindi and Hinglish: detection does as well as setting it on Hindi, Hinglish, accented Hindi and US English. Set English when the speech is only English with an Indian accent, which detection often takes for Hindi. On a subset of Svarah: 21.13% word errors with English set, 25.75% detected.
- Keywords, one per line, favour the names, places and product words your users say.
- Cactus’s Python package (cactus-needle) sends anonymous usage counts by default; set `NEEDLE_TELEMETRY=0` to turn that off. The command-line engine and the WebAssembly engine on this page make no network requests of their own.

## Credits

Every Shruti model is built on [Cactus Whistle](https://huggingface.co/Cactus-Compute/whistle) and runs in the [needle engine](https://github.com/cactus-compute/needle), both by Cactus Compute under Apache-2.0. The engine files on this page are Cactus’s WebAssembly build, unmodified, served with [its licence](https://udgamlabs.com/models/shruti/needle/LICENSE.txt) and [notice](https://udgamlabs.com/models/shruti/needle/NOTICE.txt). Shruti Bengali’s model card on Hugging Face lists its training datasets and their licences.

[Every Shruti language](https://udgamlabs.com/models/shruti#languages)

---

## About Udgam Labs

Udgam Labs is an independent AI research lab. We work from first principles: before we build, we understand why things work, whether we are training a model, fine-tuning one or applying AI to a new problem. Today that means open models such as Gargi, a 384M-parameter language model that comes within 2 points of Qwen2.5-0.5B on three reasoning benchmarks with about 1,000× less training data. Next come products, custom models for businesses, and AI in the physical world. Udgam (उद्गम) is Sanskrit for origin.

Udgam Labs · AI research from first principles · https://udgamlabs.com