We've had several tutorials with Leopard Speech-to-Text. Leopard offers fully on-device audio transcription. Today we'll build podcast transcription software using Leopard Speech-to-Text downloading new items from the RSS feed directly.
1. Backend Set up with Express.js
The backend setup will be straightforward. A single endpoint for transcription!
const express = require('express');
const app = express();
app.use(express.json());
app.use(express.urlencoded({ extended: false }));
app.use(express.static(path.join(__dirname, 'public')));
app.get('/', function (req, res) {
res.redirect('/index')
});
app.post('/rss-transcribe', async (req, res) => {
console.log("RSS feed = " + req.body.rss)
});
module.exports = app;
The test page is designed to allows entering a podcast RSS feed URL manually which will be sent to the /rss-transcribe backend for processing.
2. Parsing the RSS Feed
For parsing, use this great RSS parser from Robert Brennan. It takes the URL and provides a JSON representation of the feed:
const Parser = require('rss-parser');
app.post('/rss-transcribe', async (req, res) => {
console.log("Parsing RSS feed " + req.body.rss)
let parser = new Parser();
let feed = await parser.parseURL(req.body.rss)
console.log("Parse complete.")
}
3. Fetching the Audio
After having the JSON of the RSS, you can find where the podcast audio link was located in the object:
const fs = require('fs');
const path = require('path');
const axios = require('axios');
app.post('/rss-transcribe', async (req, res) => {
// .. parse feed
const podcastAudioUrl = feed.items[0].enclosure.url
console.log("Fetching file from " + podcastAudioUrl)
let dlResponse = await axios.get(podcastAudioUrl, { responseType: "arraybuffer" })
console.log("File obtained.")
console.log("Writing data to local file...")
const fileName = `${Math.random().toString(36).substr(2, 5)}.mp3`
fs.writeFileSync(fileName, dlResponse.data)
console.log("File write complete")
});
4. Transcribing the Podcast
Once you have the audio file, it's time to feed it into Leopard Speech-to-Text. Grab your AccessKey from the Picovoice Console for free if you haven't done yet. Replace the placeholder with your AccessKey and run the code below:
const { Leopard } = require('@picovoice/leopard-node')
app.post('/rss-transcribe', async (req, res) => {
// .. parse feed
// .. get audio file
console.log("Transcribing audio...")
const leo = new Leopard("${YOUR ACCESS KEY HERE}")
const transcript = leo.processFile(fileName)
leo.release()
fs.unlinkSync(fileName)
console.log("Transcription complete")
res.send(transcript)
});
Now, you can have the transcription being sent in the response.
What's next?
You can take it from here and enrich your solutions. Options are:
- You can write to a text file in the browser for download
- Add automation with a solution like Zapier
- Build front end with a search bar and a new endpoint for queries to make transcriptions searchable.
Resource:
Original Medium Article
Tutorial Source Code
Picovoice Leopard
Picovoice Console
Top comments (0)