In this project we are creating a newborn language from the mixture of ancient local languages such as Seto, Burgenland Croatian, Aranese, Cornish and Aromanian. We use Chat-GPT to combine words from a pair of languages and create a new original language. Using ElevenLabs we later on synthesize from text-to-speech to bring the new language to life!

Part of the KIELI project presented in Sonar 2024 (Barcelona)

The dataset is already generated in /assets/combined-words as a set of csv files. To generate a new Dataset just follow the intructions. (The prompt include the languages to guide the LLM, if you use different languages please adapt the prompt)

Using Chat-GPT 4

This file contains a set of words translated to different languages (English, Seto, Burgenland Croatian, Aranese, Cornish and Aromanian). Your mission is to create a set of csv files. Each csv file will contain the following structure: Column and row titles display one of the words in all of the languages. For each cell you will invent a new word by combining the cell column and row titles into a new word that could sound good to the ear. Create a csv file for each word. {attach csv file. find an example in assets/words-translated.csv}

Ask Chat-GPT to export the csv files as a zip file

This repo uses ElevenLabs API. Please create an account to get an API Key. Create a .env file and copy your API Key

ELEVEN_LABS_KEY=***********

https://github.com/pyenv/pyenv

The python verison used in this repo is specified in .python-version

Activate venv before installing requirements. source venv/bin/activate

pip install requirements.txt

Activate venv before running the script. source venv/bin/activate

python generate_speech.py