Use multiple GPU's

[ehnwebmaster]

vertd only uses the first GPU available:

For example .mp4 to .mkv command:

ffmpeg -hide_banner -loglevel error -progress pipe:1 -hwaccel cuda -i input/file.mp4 -c:v h264_nvenc -c:a aac -strict experimental -preset medium -b:v 5058664 -map_metadata 0 -map_chapters 0 output/file.mkv

Need to use -gpu n parameter

ffmpeg -i video1.mp4 -c:v h264_nvenc -gpu 0 output1.mkv

Since you can't use multiple gpu's to encode a video use a queue something like this will be works:

ffmpeg -i video2.mp4 -c:v h264_nvenc -gpu 2 output2.mkv
ffmpeg -i video3.mp4 -c:v h264_nvenc -gpu 3 output3.mkv

Use one GPU per video

Rust Example:

use std::sync::atomic::{AtomicUsize, Ordering};
use std::process::Command;

// Variable global o en el estado de tu aplicación (State<T>)
static JOB_COUNTER: AtomicUsize = AtomicUsize::new(0);

fn convertir_video(input: &str, output: &str) {
    // 1. Obtenemos el número de trabajo actual
    let current_job = JOB_COUNTER.fetch_add(1, Ordering::SeqCst);
    
    // 2. Calculamos qué GPU usar (Módulo 3: devuelve 0, 1 o 2 rotativamente)
    let gpu_id = current_job % 3;

    println!("Asignando trabajo {} a la GPU {}", current_job, gpu_id);

    // 3. Construimos el comando
    let status = Command::new("ffmpeg")
        .arg("-i").arg(input)
        .arg("-c:v").arg("h264_nvenc")
        
        // AQUÍ ESTÁ LA MAGIA:
        .arg("-gpu").arg(gpu_id.to_string()) 
        
        .arg("-c:a").arg("aac")
        .arg(output)
        .status()
        .expect("Falló ffmpeg");
}