Capturing audio from chromium in dockerized headless mode?

[princefishthrower]

First off, thanks for this library - it's been amazing to use!

I've finally dockerized my workflow and my .webms are being produced just fine - the video is high quality and looks fine - but there is no sound.

I realize this isn't a question directly related to the library, but I'm wondering how behind the scenes puppeteer-stream finds audio... does anyone have an example of a dockerized version that is able to correctly capture audio from a headless chromium? Currently I'm using FROM alpine:latest but I'm worried I may have to go full OS (ubuntu or similar) to get the audio working in a headless scenario. If it helps, my puppeteer browser call is like so:

const browser = await launch({
        dumpio: true,
        // macOS:
        // executablePath: "/Applications/Google Chrome.app/Contents/MacOS/Google Chrome",
        // linux (docker):
        executablePath: "/usr/bin/chromium-browser",
        headless: "new", // supports audio!
        // headless: false, // for debugging
        defaultViewport: { width: 1920, height: 1080 },
        args: [
            '--window-size=1920,1080',
            '--start-fullscreen',
            '--ozone-override-screen-size=1920,1080', // for linux
            '--no-sandbox', // to run as root on docker
            '--allow-file-access-from-files',
            '--enable-audio-service',
            '--mute-audio=false',
        ]
    });

I know audio capture works with headless because before I dockerized everything sound was working fine! :)

[SamuelScheit]

Yeah it could be because of alpine and missing audio libraries.
You could try installing them manually

RUN apk add --no-cache  alsa-lib alsa-utils ffmpeg

or try for testing purposes ubuntu as a base image.

[princefishthrower]

Thanks for the quick reply, unfortunately, I'm still getting errors whenever code like this is called in my chromium:

const audioElement = new Audio(mp3Url);
audioElement.volume = volume;
audioElement.onended = () => {
  resolve();
};
audioElement.onerror = (error) => {
  console.error(error);
};
audioElement.play().catch(async error => {
    console.error(error);
});

Puppeteer isn't able to deserialize those errors very well, so I just see JSHandle@object... but I'm determined to figure out what the heck is missing so I'll report back once I get audio working.

[SamuelScheit]

Audio works for me with this dockerfile

FROM zenika/alpine-chrome:with-puppeteer

COPY . .

CMD ["node", "examples/file.js"]

[princefishthrower]

Crazy question, but what operating system does that work on? Linux? At first, I was concerned because I am doing new Audio(<my mp3 URL here>) instead of directly recording a site like YouTube (your example)... but I'm pretty sure it's not the root issue because my audio elements work find when I test from localhost directly in a browser tab...

But I'm on an M3 and wondering if it's the typical "docker promises OS independence but secretly not really" 😄

[SamuelScheit]

I was testing on macOS 15.3.1 running Docker 4.39.0

[princefishthrower]

Is it mac silicon or Intel?

[SamuelScheit]

I've now tested on silicon and intel, it works on both

see this commit 9d133e3