Funky Talker

Funky Talker is a small Flask text-to-speech prototype that uses local Piper voice models. Enter text in the web page, choose a voice, submit the form, and the app generates a WAV file that can be played back in the browser.

The app can run either directly with local Python or inside Docker. The Docker image includes Python, Flask, and Piper; the voice models stay outside the image and are mounted into the container at runtime.

Current Features

• Flask web interface for entering text
• Voice selection from locally installed Piper .onnx models
• Browser playback of generated WAV audio
• Configurable Piper command through PIPER_COMMAND
• Docker image build and run workflow
• Local model files kept out of git

Project Structure

funky_talker/
|-- Dockerfile
|-- requirements.txt
|-- funky_talker.py
|-- templates/
|   `-- index.html
|-- static/
|   `-- output.wav        # generated at runtime
|-- models/               # local Piper models, not committed
|-- input.txt             # generated at runtime
|-- .dockerignore
`-- .gitignore

Requirements

For local Python:

• Python 3
• Flask
• Piper text-to-speech
• One or more Piper voice models

For Docker:

• Docker Desktop
• One or more Piper voice models

Voice models are available from:

https://huggingface.co/rhasspy/piper-voices

Voice Models

Voice models are not included in this repository because they are large binary files. Each Piper voice usually has two files:

voice-name.onnx
voice-name.onnx.json

Place both files in the local models/ folder.

The current app expects these model files:

models/en_US-lessac-medium.onnx
models/en_US-lessac-medium.onnx.json

models/en_US-lessac-high.onnx
models/en_US-lessac-high.onnx.json

models/en_GB-northern_english_male-medium.onnx
models/en_GB-northern_english_male-medium.onnx.json

models/en_US-john-medium.onnx
models/en_US-john-medium.onnx.json

models/en_US-kristin-medium.onnx
models/en_US-kristin-medium.onnx.json

To add another voice, download the .onnx and .onnx.json files, place them in models/, then add an entry to the VOICE_MODELS dictionary in funky_talker.py.

Local Python Setup

Create and activate a virtual environment:

python -m venv .venv
.\.venv\Scripts\Activate.ps1

Install dependencies:

pip install -r requirements.txt

If Piper is not on your system PATH, set PIPER_COMMAND before starting the app:

$env:PIPER_COMMAND = "C:\path\to\piper\piper.exe"

Run the app:

python funky_talker.py

Then open:

http://127.0.0.1:5000

Docker Setup

Build the image from the project root:

docker build -t funky-talker .

Run the container and mount the local models/ directory:

docker run --rm -p 5000:5000 -v ${PWD}/models:/app/models funky-talker

Then open:

http://localhost:5000

On Windows PowerShell, ${PWD} should resolve to the current project folder. If Docker has trouble with that path, use the full path instead:

docker run --rm -p 5000:5000 -v D:\personal\code\funky_talker\models:/app/models funky-talker

Generated Files

The app creates a few local runtime files:

• input.txt
• static/output.wav

These are generated output and should not be committed.

Portfolio Notes

This project demonstrates:

• Python application structure
• Flask request/response handling
• Template rendering with Jinja
• Calling a local command-line AI/TTS tool from Python
• Managing local model assets outside git
• Docker image creation and container runtime usage

Possible next improvements:

• Move model definitions into a config file
• Add character or speaking-style presets
• Improve the visual design of the web interface
• Add a compose.yaml for a shorter Docker startup command
• Add basic error handling when a model file is missing