Skip to content

About

speech device

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

A useful tool to help correct stuttering

Dependencies

This project is built off of this github repo which has been massively helpful

please install these:

mkdir demo_models && cd demo_models
mkdir asr && cd asr
gdown --id 1BeT7m_5qv19Sb5yrZ2zhKu6fEprUoB9N -O config.json
gdown --id 13n8VrTFVq4jGouCDamkReHlHm_1yz20U -O pytorch_model.bin
cd ..
gdown --id 1GQIXgCSF3Usiuy5hkxgOl483RPX3f_SX -O language.pt
gdown --id 1wWrmopvvdhlBw-cL7EDyih9zn_IJu5Wr -O acoustic.pt
gdown --id 1LPchbScA_cuFx1XoNxpFCYZfGoJCfWao -O multimodal.pt

The following packages are needed:

  • pandas==1.5.0
  • torch==1.12.1
  • torchaudio==0.12.1
  • transformers==4.22.2
  • whisper_timestamped==1.14.4
  • gdown==5.1.0

How to use

It's really simple: this stuttering-correction tool runs completely in the terminal. Create a Python virtual environment, install the dependencies, and start speaking about a topic you want to discuss. The tool then runs a 2-minute conversation, analyzes your speech, and displays words that may need practice so you can repeat them. This exercise can help improve speech disorders, especially stuttering.

Future Plans

Although this only runs in a terminal, a GUI browser version is currently being developed and will be released later this month!

Frontend Prototype

A browser-based frontend is now included in `frontend/`.

Run it locally with one unified web server:

python backend_api.py

Then open http://localhost:8787 in your browser.

If you see a startup SyntaxError, pull the latest code and rerun python backend_api.py (the server file should start with comment lines, not an open triple-quoted block).

To expose this on the web, run on any host/VM/Raspberry Pi with a public DNS/IP and open the same port:

python backend_api.py --host 0.0.0.0 --port 8787

The frontend and API are served together from one process, and the frontend uses relative /api/... paths so it works on any domain without hard-coded localhost ports. Run it locally with:

cd frontend
python -m http.server 4173

Then open http://localhost:4173 in your browser.

The UI is now redesigned for a vertical Raspberry Pi touchscreen layout (optimized around ~480x800) with a light blue + white visual style tuned for clarity.

Frontend Screens

Current simplified screens in this prototype: - Live transcript screen (tap **Start Conversation** to begin) - Live transcript screen - Loading / analysis screen - Final results screen (practice words + summary)

derek thingy

Setuptools 68.2.2 pip install --no-build-isolation -r requirements.txt

IN POWERSHELL:

Do this too:

Step A — Recreate with Python 3.10 py -3.10 -m venv venv

Step B — Fix execution policy Run: Set-ExecutionPolicy -Scope Process -ExecutionPolicy Bypass This is temporary (safe) and only applies to this terminal.

Step C — Activate venv .\venv\Scripts\Activate.ps1

About

speech device

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages