Duyu-small

Duyu-small

Transcribes Turkish and English speech inside your browser from a recording or an uploaded file. A small model runs on your own device, so your audio stays on it.

Made by KaanAydinli on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: Duyu-small is a Turkish and English speech recognizer that runs inside a web page. You record or upload audio, and the model turns it into text on your own device. It uses onnxruntime-web, so nothing is sent to a server.

Who it suits: People who need Turkish or English speech turned into text and want the audio to stay on their device, and developers who want an example of a speech model running in a browser.

Industry | Who uses it | How they use it | Example
Language work | Turkish and English speakers | Get text from a short spoken recording | Record a sentence and read the text
Media | Writers and editors | Draft notes from an audio file without uploading it | Upload an interview clip and read the draft
Software | Web developers | Study speech recognition that runs in the browser | Read engine.js and worker.js
Research | Speech researchers | Time the model on their own processor | Read the speed figure the page reports

Anyone who needs a language other than Turkish or English can skip it.

Where to use it: It runs in these places.

• A web browser, on the Space's page at Hugging Face
• Windows, macOS or Linux, by opening the downloaded index.html in a browser

Use case | Who | What they get back
Transcribe a recording | Anyone with a microphone | Text from what you said, made on your device
Transcribe an uploaded file | Editors and writers | Text from the file, with the audio kept local
Check the speed | Researchers | A report of how fast the model ran on your processor

How to set it up: Using the Space online needs only a browser. Getting its files onto your own computer needs the following.

• Python and pip, for the huggingface_hub package that gives you the hf command
• A web browser
• Disk space for the repository, which holds 9 files

1. Install the huggingface_hub package.
2. Download the Space's files into a local folder.
3. Move into that folder.
4. Open the file index.html in a web browser.

pip install -U huggingface_hub
hf download KaanAydinli/Duyu-small-demo --repo-type space --local-dir Duyu-small-demo
cd Duyu-small-demo

How to use it: Open the page, then record or upload audio. The model downloads once and the browser caches it, so later visits skip that wait. The page reports how fast it ran on your CPU.

Tool | What it does
index.html | The page you open in a browser
engine.js | The speech engine, ported from the model's Python version
worker.js | Runs the work in a browser worker
vendor | Holds the libraries the page needs

Example: You record yourself reading a short Turkish sentence and stop the recording. The page returns the text of what you said, and it reports how fast that run was on your processor.

For newcomers and experts: Speech recognition means turning spoken words into written text. The first small thing to try is the online page with a short recording in Turkish or English.

The model has 19.3 M active parameters and comes in int8 and 4-bit ONNX forms. The engine ports the Python one line by line. It computes log-mel features, makes three shared encoder passes, and runs an attention beam search with CTC rescoring, plus guards for silence and runaway output.

You record or upload audio
The page computes log-mel features
Three shared encoder passes run
Beam search with CTC rescoring gives the text

Good to know: The model needs a first download before it can run, and the browser then keeps it.

Title: Model download size
int8: 56 MB
4-bit: 34 MB

• License: apache-2
• Cost to run: none named
• Data it reaches: the model download, plus the audio you record or upload, which stays on your device
• Not for: languages other than Turkish and English

On Hugging Face, 13 November 2026

• Likes: 2
• Built with: static
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/KaanAydinli/Duyu-small-demo
• Files: https://huggingface.co/spaces/KaanAydinli/Duyu-small-demo/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.