How to run ffmpeg.wasm directly on the main thread (usacase: cloudflare worker, browser extension service worker, etc)

[clmnin]

I am trying to run ffmpeg wasm on a cloudflare worker with a custom build. But since I can't start a web worker from within a CF worker I get the error

Caution

Worker is not defined

Is there a way to run ffmpeg.wasm on the main thread (in my case directly in the CF worker)?

Running ./build-with-docker.sh to build a very small version of ffmpeg.wasm. By following this blog post - link. The final wasm file is only 8.4 mb (2.8 mb gzipped). CF Worker can support upto 10 MB compressed.

Here is the code for the worker

import { FFmpeg } from '@ffmpeg/ffmpeg';
import { fetchFile } from '@ffmpeg/util';

export interface Env {
	// Add any environment variables here
}

export default {
	async fetch(request: Request, env: Env, ctx: ExecutionContext): Promise<Response> {
		if (request.method !== 'POST') {
			return new Response('Send a POST request with a video file', { status: 400 });
		}

		const formData = await request.formData();
		const videoFile = formData.get('video') as File | null;

		if (!videoFile) {
			return new Response('No video file provided', { status: 400 });
		}

		try {
			const trimmedVideo = await trimVideo(videoFile);
			if (!trimmedVideo) {
				return new Response(`Failed to trim video`, { status: 500 });
			}
			return new Response(trimmedVideo, {
				headers: {
					'Content-Type': 'video/mp4',
					'Content-Disposition': `attachment; filename="trimmed_${videoFile.name}"`,
				},
			});
		} catch (error) {
			// @ts-ignore
			return new Response(`Error processing video: ${error.message}`, { status: 500 });
		}
	},
};

async function trimVideo(videoFile: File): Promise<Uint8Array | null> {
	const ffmpeg = new FFmpeg();
	try {
		await ffmpeg.load({
			coreURL: 'http://localhost:8787/ffmpeg-core.js',
			wasmURL: 'http://localhost:8787/ffmpeg-core.wasm',
		});
	} catch (error) {
		// @ts-ignore
		console.error(error.message);
		return null;
	}

	const inputFileName = 'input.mp4';
	const outputFileName = 'output.mp4';

	ffmpeg.writeFile(inputFileName, await fetchFile(videoFile));

	await ffmpeg.exec([
		'-i',
		inputFileName,
		'-ss',
		'1', // Start at 1 second
		'-to',
		'-1', // End 1 second before the end
		'-c',
		'copy', // Use the same codec (faster)
		outputFileName,
	]);

	const data = await ffmpeg.readFile(outputFileName);
	await ffmpeg.deleteFile(inputFileName);
	await ffmpeg.deleteFile(outputFileName);

	return new Uint8Array(data as ArrayBuffer);
}

[francosalcedo]

any news?

[clmnin]

Nope. This project is in the back burner for now. May come back to it in a new months.

A few ideas to explore are

1. WASI
2. Building wasm for CF worker

Interesting links:

1. https://github.com/SebastiaanYN/FFmpeg-WASI
2. https://github.com/Yahweasel/libav.js

Note: To make ffmpeg work on a CF worker you probably need a paid account which can support up to 10 MB workers

[quink-black]

I have added some wasm optimization to FFmpeg
https://github.com/FFmpeg/FFmpeg/commits?author=quink-black

And it's easy to build and run upstream FFmpeg with wasi
https://github.com/quink-black/ffmpeg-ci/blob/master/wasi_ffmpeg.sh