Build v86 for one program/purpose: TTS

[guest271314]

How to build this containing a single program: eSpeak NG?

Why: Web Speech API is broken; extra work is needed for speak() to work on Firefox; neither Chrome nor Firefox support SSML input, nor capturing audio output of the local speech synthesis engine.

I don't need everything else, just the capability to run the eSpeak NG programm and get --stdout which I can parse to create a live MediaStreamTrack from in the browser.

Thanks in advance for any help with this.

[cemalgnlts]

Yes, I'm wondering if anyone has done this too.

I'm building a linux with buildroot just to run nodejs: cemalgnlts/now-linux-build

Instead of recompiling Nodejs, I'm using the prebuilt binary so it takes less time to compile I've shrunk busybox to only support a few commands.

Without Nodejs it's ~4MB but not small enough. I'm sure there are many more files that Nodejs doesn't need.

[guest271314]

I would suggest experimenting with QuickJS qjs which is 5MB, less than 1mb after strip unless you have a compelling need to use node executable.

[cemalgnlts]

I also looked at it as an alternative but I want Node projects to run directly but for QuickJS also Node APIs should be written which takes a lot of effort.

[guest271314]

Why do you need Node.js API's?

[guest271314]

Note, it is possible to run node controlled from the browser using Native Messaging https://github.com/guest271314/native-messaging-nodejs.

[cemalgnlts]

Yes, I've looked at that before. I'm building a project more like WebContainers - Nodebox.

[guest271314]

Good luck.

[cemalgnlts]

Thanks.

[guest271314]

Browsing through your source code it looks like embedding eSpeak NG in buildroot should work.

Did you use Rust at all during the build process?

[guest271314]

The Linux build was not built specifically for this project, so there are unnecessary files.

That's what I am trying to avoid.

[cemalgnlts]

You need to remove the addons for Node:

build.sh
inittab

Then run the ./build.sh command.

When the process is complete the linux image will be in this folder: buildroot-2023.02/output/images/rootfs.iso9660

I am using https://buildroot.org to build the Linux image.

I've been looking for a way to make it even smaller, but more is beyond me.

I started building the project from this repo: https://github.com/humphd/browser-vm.

[guest271314]

The more I keep looking at this the more I realize the working solution I have with Native Messaging is probably the best fit for me.

I don't want to create a duplicate copy of a program to carry around for use in the browser when I already have the program installed on the machine.

I can just run node, deno, bun, qjs, tjs, and espeak-ng from the browser - without compiling a duplicate version of those programs to run in the browser.

I was looking at your code here [scripts/boot.js](https://github.com/cemalgnlts/now-linux-build/blob/master/scripts/boot.js) and it looks like you are using Node.js to build the program, then you download node again, just to use in the browser.

[cemalgnlts]

v86 allows you to keep the current state while linux is running. I build the linux image using CircleCI(.circleci/config.yml) (GitHub Actions alternative), then I run v86 in the node environment and save the current state when the linux boot process is finished. In this way, I run it from the state file I saved instead of linux.iso and I don't have to wait for the boot process every time.

Anyone can run Node on their device, but embedding it in the browser will make it more accessible, interactive documents etc.

[guest271314]

Anyone can run Node on their device, but embedding it in the browser will make it more accessible, interactive documents etc.

I can access the node executable I download from the browser, run any program I want using that executable and pipe the stdout from that program or node itself to the browser. So I don't need to create a duplicate node to embed in the browser.

If the node you are embedding in the browser cannot access the local file system or run programs then it is a less than complete node embedment.

I am just trying to justify continuing to pursue compiling or porting to WASM and I keep encountering the fact that I will be creating a duplicate espeak-ng compilation, when I can already use the espeak-ng (or node) I have installed or accessible (I never actually install node, I just use the nightly node executable https://github.com/guest271314/download-node-nightly-executable).