Open a shell script and create a new python environment.
conda create --name translator python=3.9
conda activate translator
git clone https://github.com/stephanj/translation_transcript_service.git
Install the required Python modules
cd translation_transcript_service/src/main/server
pip3 install -r requirements.txt
Set your OpenAI key in .env
echo "OPENAI_API_KEY='sk-xxxxxxxxxxxxxxxxxxxxxxxxxxxx'" > .env
Get your OpenAI key from https://platform.openai.com/account/api-keys
For the summary audio you need to have an ElevenLabs key, but this is optional.
https://beta.elevenlabs.io/speech-synthesis click on the profile menu itme.
Add this key in the config.js file.
Create a config.js file in src/main/webapp/js which holds your ElevenLabs key. This is used to convert the transcript summary and turn it into audio which is played on the Summary slide.
const config = {
audioKey: '0e5cbxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx'
};You need to update the exported HTML slides to include the translation buttons and related javascript code.
In the generated index.html file add the following:
The css for the buttons and translation text layout
<link rel="stylesheet" href="../../webapp/css/styles.css"> The actual buttons in the header. Place this right underneath the body tag.
<body id="body" bgcolor="black">
<div class="container">
<div id="translatedText"></div>
<div class="filler">
<label for="source-lang">From</label>
<select id="source-lang">
<option value="en">English</option>
<option value="nl">Dutch</option>
<option value="fr">French</option>
<option value="es">Spanish</option>
<option value="gr">Greek</option>
<option value="it">Italian</option>
<option value="ja">Japanese</option>
<option value="zh">Chinese</option>
<option value="ar">Arabic</option>
</select>
<label for="target-lang">To</label>
<select id="target-lang">
<option value="fr">French</option>
<option value="nl">Dutch</option>
<option value="es">Spanish</option>
<option value="en">English</option>
<option value="gr">Greek</option>
<option value="it">Italian</option>
<option value="ja">Japanese</option>
<option value="zh">Chinese</option>
<option value="ar">Arabic</option>
</select>
<button id="start">Start</button>
<button id="stop" disabled>Stop</button>
</div>
</div> And at the end of the HTML tag add the needed javascript import.
<script src="assets/player/main.js"></script> <!-- already there -->
<script src="../../webapp/js/config.js"></script> <!-- the config with the elevenLabs key -->
<script src="../../webapp/js/translator.js"></script> <!-- the translator script -->I don't use PowerPoint, so maybe someone else can do this?
- Start the websocket python app which will accept audio chuncks from the browser.
python3 websocket.py
- Open the updated index.html which contains the slides.
- Select the from/to languages and press "Start".
Every 20 seconds the audio is pushed to the python app, it converts the webm to text using Whisper and asks ChatGPT to translate it.