# OmniGet vs aria2: an aria2 alternative with a window, timed

- URL: https://getomniget.com/benchmark/omniget-vs-aria2
- Updated: 2026-09-27
- OmniGet 1.0.0

> Free, open-source yt-dlp GUI for Windows, macOS and Linux. Paste a link to save video, audio and files from 1,000+ sites, or run AI coding agents.

---

[Benchmarks](https://getomniget.com/benchmark)

# OmniGet vs aria2

OmniGet is a free, open-source (GPL-3.0) desktop app for Windows, macOS and Linux that uses aria2 as one of its engines. Timed on September 27, 2026, aria2 1.37.0 and the OmniGet 0.10.1 app tied on a 372 MB direct file (80.5 s against 80.1 s) and on a 276 MB torrent from a magnet link (70.9 s with the same trackers against 70.1 s); aria2 used a fifth of the memory, and OmniGet also downloads from site pages, which aria2 cannot read.

Measured September 27, 2026 · 43.5k stars · Latest release release-1.37.0 · – views · [aria2 on GitHub](https://github.com/aria2/aria2)

Written by [Antonio Leandro](https://getomniget.com/about), who makes OmniGet. aria2 is described from its own repository and releases.

## Short answer

On this line of about 37 Mbps, aria2 and OmniGet ran at the same speed: the 372 MB file took about 80 seconds with either, and a 276 MB torrent about 70 seconds once aria2 had the same public trackers OmniGet adds. aria2's -x16 -s16 -k1M flags saved about 3 seconds on the file, inside the spread between rounds. aria2 is the lighter tool (18 MB of memory against 88 MB), stops seeding at a ratio of 1.0 and runs as a server with an RPC interface. OmniGet downloaded all 9 free test items, aria2 the 2 that were plain links (the file and the torrent): OmniGet reads video pages and HLS streams, which aria2 cannot, and puts files and torrents in one window.

## Pros and cons, side by side

Written from our study of each app's repository and releases. Every row names the better app, including when it is not OmniGet.

| What matters | OmniGet | aria2 |
| --- | --- | --- |
| Speed on a large direct file | 80.1 s for 372 MB (79.2 to 82.6 s) | 80.5 s as installed, 77.4 s with -x16 -s16 -k1M; the ranges overlap |
| Speed on a torrent | 70.1 s for 276 MB from a magnet (64.0 to 109.4 s) | 82.2 s as installed, 70.9 s with the same public trackers; the ranges overlap |
| What it downloads | 9 of 9 free test items: video pages through yt-dlp and its own extractors, HLS, torrents, direct files (better) | 2 of 9, the direct file and the torrent: links you already have over HTTP(S), FTP, SFTP, BitTorrent or Metalink |
| Getting started | Installer for Windows, macOS and Linux; paste a link in a window (better) | Terminal flags; official builds only for Windows and Android, macOS through Homebrew |
| After a torrent finishes | Keeps seeding with no ratio, time or upload limit until you remove it | Stops at share ratio 1.0 by default; ratio, time and upload limits (better) |
| Scripts, servers and footprint | Desktop app and MCP server; 88 MB of memory at the peak; command line without torrents | Daemon, JSON-RPC over WebSocket, input files, hooks; 2.5 MB zip, 18 MB of memory (better) |

OmniGet wins 2, aria2 wins 2, 2 tied

## Download speed

Direct file from download.blender.org (372 MB), as installed (Lower is better · s):

- OmniGet: 80.1 s
- aria2: 80.5 s

Big Buck Bunny torrent from a magnet link (276 MB), as installed (Lower is better · s):

- OmniGet: 70.1 s
- aria2: 82.2 s

## What each app can do

|  | OmniGet | aria2 |
| --- | --- | --- |
| Playlists and channels | Yes | No |
| Subtitles | Yes | No |
| Audio only | Yes | No |
| Choose quality and format | Yes | No |
| Private and members-only videos (your login) | Yes | Partly |
| Browser extension | Yes | No |
| Torrents and magnet links | Yes | Yes |
| Direct file links | Yes | Yes |
| Queue and parallel downloads | Yes | Yes |
| Watch channels for new videos | Yes | No |
| SponsorBlock, chapters, metadata | Yes | No |
| AI agents and MCP | Yes | No |

### After the download: seeding and servers

Left alone, aria2 keeps seeding until it has uploaded as much as it downloaded and then exits; you can set a ratio, a time or an upload limit. OmniGet 0.10.1 seeds with no limit until you remove the torrent or quit the app. aria2 also runs as a daemon that other programs drive over JSON-RPC or WebSocket, which is how web front ends such as AriaNg work. OmniGet's equivalent for automation is its MCP server, which lets an AI agent queue a magnet link or a video URL.

## Who wins what

-   Direct file from download.blender.org (372 MB), as installed

    OmniGet

    OmniGet 80.1 s · aria2 80.5 s

-   Big Buck Bunny torrent from a magnet link (276 MB), as installed

    OmniGet

    OmniGet 70.1 s · aria2 82.2 s

aria2 is a command-line download utility in C++ for HTTP(S), FTP, SFTP, BitTorrent and Metalink, known for opening many connections per file. It downloads a URL, a torrent or a Metalink you already have; it does not read web pages. OmniGet is a desktop app that finds the file on a page through yt-dlp and its own extractors, downloads direct files with its own segmented downloader and torrents with its own client, and hands plain files to aria2 inside yt-dlp when aria2c is installed (on Windows it downloads aria2c by itself).

aria2 1.37.0 (November 2023, still the latest release) from Homebrew was tested twice, as installed and with the speed flags its manual documents, against the published OmniGet 0.10.1 app, on the same Mac, network and day. The direct file is Tears of Steel from download.blender.org; the torrent is WebTorrent's free Big Buck Bunny, as a magnet link without its web seed so both tools download from the swarm only. Each item had a warm-up round and five timed rounds in alternating order.

## Download size and packages

Free test items downloaded correctly (9 in one pass) (Higher is better · %):

- OmniGet: 100 %
- aria2: 22 %

Peak memory while downloading the 372 MB file (Lower is better · MB):

- aria2: 17.8 MB
- OmniGet: 88 MB

Peak CPU while downloading the 372 MB file (100% = one core) (Lower is better · %):

- aria2: 12.4 %
- OmniGet: 85.7 %

### Memory, CPU and install

At the peak on the direct file, aria2 used 18 MB of memory and 12% of a core; the OmniGet app used 88 MB and 86% of a core for its main process, without counting its window. aria2 ships official builds only for Windows (a 2.5 MB zip) and Android; on a Mac it comes from Homebrew, and the last official macOS build was 1.35.0 in 2019. OmniGet has installers for Windows, macOS and Linux.

[Download OmniGet for Windows, macOS or Linux](https://getomniget.com/download)

## The apps in this benchmark

-   ### [OmniGet](https://getomniget.com)

    - License: GPL-3.0
    - Built with: Tauri (Rust + web UI)
    - GitHub: 14.4k stars · v0.10.1

    -   Its own extractors, torrents and direct files on top of yt-dlp.
    -   An MCP server and AI coding agents in the same app.
-   ### [aria2](https://github.com/aria2/aria2)

    - License: GPL-2.0
    - GitHub: 43.5k stars · release-1.37.0

## The direct file: the line is the limit

aria2 as installed opens one connection per server (-x defaults to 1), yet it finished the 372 MB file in 80.5 seconds, the same as the OmniGet app with its segmented downloader (80.1 s) and as aria2 with 16 connections (77.4 s). Every run was close to the line speed of about 37 Mbps, so more connections had nothing to win. On a faster line or a server that throttles each connection, aria2's -x16 and OmniGet's segments would pull ahead of a single connection; this test does not show that case. aria2 caps connections per server at 16, a limit people have asked to raise since 2017 (issue 1039).

## The torrent: trackers matter more than the client

From the magnet link, aria2 as installed took 82.2 seconds (median) and the OmniGet app 70.1. The magnet names only two trackers; OmniGet adds 10 public trackers to every torrent by default. Given the same 10 trackers, aria2 took 70.9 seconds, and the rounds of all three overlapped (64 to 109 seconds), so the swarm set the pace. Both stop the clock at the last piece here: aria2 was run with --seed-time=0, and OmniGet was stopped when it started seeding.

## Which one should you pick?

-   **Choose OmniGet if** you want to paste a page link and get the video, audio or file, keep torrents and downloads in one window, or let an AI agent queue downloads, without writing aria2 flags.
-   **Choose aria2 if** you already have the file URLs, torrents or Metalinks, need FTP, SFTP, mirrors or checksums, want seeding limits, or run downloads on a server through its RPC interface.

## Other aria2 alternatives

Free, open-source apps that do some of what aria2 does.

-   [Open-source yt-dlp GUIs compared](https://getomniget.com/benchmark/yt-dlp-gui) Every desktop app we track, ordered by GitHub stars, with its license, latest release and who it suits.
-   [OmniGet vs ezytdl](https://getomniget.com/benchmark/omniget-vs-ezytdl) Electron front end for yt-dlp with advanced options. Compared with OmniGet
-   [OmniGet vs Media Downloader](https://getomniget.com/benchmark/omniget-vs-media-downloader) Qt/C++ front end for several download engines, including yt-dlp, gallery-dl and aria2c. Compared with OmniGet
-   [OmniGet vs SCrawler](https://getomniget.com/benchmark/omniget-vs-scrawler) Windows app that saves media from social sites and user profiles. Compared with OmniGet
-   [OmniGet vs Videomass](https://getomniget.com/benchmark/omniget-vs-videomass) Cross-platform FFmpeg front end for batch conversion with presets; it removed its yt-dlp downloader in version 6.1.12. Compared with OmniGet
-   [OmniGet vs Vividl](https://getomniget.com/benchmark/omniget-vs-vividl) Windows front end for yt-dlp built with WPF. Compared with OmniGet
-   [OmniGet vs Hindsight](https://getomniget.com/benchmark/omniget-vs-hindsight) Agent memory system by Vectorize: retain, recall and reflect over a memory bank, as a server with Postgres or embedded in Python. Compared with OmniGet
-   [OmniGet vs you-get](https://getomniget.com/benchmark/omniget-vs-you-get) Command-line downloader written in Python that scrapes media from web pages. Compared with OmniGet
-   [OmniGet vs youtube-dl](https://getomniget.com/benchmark/omniget-vs-youtube-dl) The original command-line video downloader that yt-dlp was forked from. Compared with OmniGet
-   [OmniGet vs yt-dlp](https://getomniget.com/benchmark/omniget-vs-yt-dlp) The command-line audio and video downloader that OmniGet runs on. Compared with OmniGet
-   [OmniGet vs Open Video Downloader](https://getomniget.com/benchmark/omniget-vs-open-video-downloader) Cross-platform yt-dlp front end built with Tauri and Vue, with a very small installer. Compared with OmniGet
-   [OmniGet vs Parabolic](https://getomniget.com/benchmark/omniget-vs-parabolic) yt-dlp front end with a native interface: WinUI on Windows and GNOME on Linux, published on Flathub. Compared with OmniGet
-   [OmniGet vs VidBee](https://getomniget.com/benchmark/omniget-vs-vidbee) Electron app built on yt-dlp that also makes local transcripts, AI summaries and RSS subscriptions. Compared with OmniGet
-   [OmniGet vs ytDownloader](https://getomniget.com/benchmark/omniget-vs-ytdownloader) Electron yt-dlp app with a clip trimmer and a video compressor, on many app stores. Compared with OmniGet
-   [yt-dlp on GitHub](https://github.com/yt-dlp/yt-dlp) The command-line audio and video downloader that OmniGet runs on.
-   [Motrix on GitHub](https://github.com/agalwood/Motrix) Desktop download manager built on aria2, with HTTP, BitTorrent and magnet links.
-   [Persepolis on GitHub](https://github.com/persepolisdm/persepolis) Download manager written in Python that drives aria2, with scheduling.

## Where OmniGet still loses

We publish this on purpose. It is the list we work from to make OmniGet better.

-   Seeding control: OmniGet 0.10.1 keeps seeding a finished torrent until you remove it or quit, with no ratio, time or upload limit. aria2 stops at a share ratio of 1.0 by default and has --seed-ratio, --seed-time and upload limits.
-   Torrent features: aria2 uses web seeds (BEP 19), encrypted peer connections and Local Peer Discovery; OmniGet's torrent client (librqbit 8.1.1) has none of the three.
-   Protocols and sources: aria2 also speaks FTP, SFTP and Metalink, checks a --checksum, and pulls one file from several mirrors and the swarm at once. OmniGet takes one HTTP(S) URL or a torrent.
-   Scripts and servers: aria2 runs as a daemon with JSON-RPC and XML-RPC over HTTP and WebSocket, reads URLs from a file or stdin, runs hooks on completion and saves its session to resume later. OmniGet's command line has no torrents, and seeding is not restored after a restart.
-   Footprint: aria2 is a 2.5 MB zip for 32- and 64-bit Windows and Android, and used about 18 MB of memory here, against 88 MB for the OmniGet app.

## Trying OmniGet costs nothing

Both save normal files, so nothing needs migrating. OmniGet uses aria2c for plain files inside yt-dlp when it finds it (and installs it by itself on Windows), so you can keep aria2 for scripts and servers and use OmniGet for pages and a queue.

## Method

Run it yourself

`ARIA2C=<aria2c 1.37.0> node scripts/bench/speed.mjs --slug omniget-vs-aria2 --tools omniget-app,omniget-cli,aria2,aria2-tuned --items direct-large --speed-only --rounds 5 --warmup 1 --startup; then the same with --tools omniget-app,aria2,aria2-tuned --items torrent` 

[Raw data (JSON)](https://getomniget.com/benchmark/omniget-vs-aria2.json)

Data licensed [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/): reuse it freely with a link to this page.

### Sources

-   [github.com/tonhowtf/omniget](https://github.com/tonhowtf/omniget)
-   [getomniget.com](https://getomniget.com)
-   [github.com/aria2/aria2](https://github.com/aria2/aria2)

## Questions

Is aria2 still maintained?

Slowly. The latest release is 1.37.0 from November 15, 2023; commits kept landing on the main branch into 2026, but no new release has shipped. The Windows and Android builds of 1.37.0 are the current official ones.

Does OmniGet use aria2?

Yes, in one place: when aria2c is installed, OmniGet has yt-dlp hand plain HTTP and FTP files to it with 8 to 16 connections, and keeps HLS and DASH streams on yt-dlp's own downloader. On Windows OmniGet downloads aria2c by itself. Direct links pasted into OmniGet go to its own segmented downloader, and torrents to its own client.

Is there a GUI for aria2?

aria2 has no official one. Front ends such as AriaNg (in a browser), Motrix and Persepolis drive it through its RPC interface. OmniGet is not an aria2 front end: it has its own downloader and torrent client, and adds site pages. See [OmniGet vs Motrix](https://getomniget.com/benchmark/omniget-vs-motrix) and [OmniGet vs Persepolis](https://getomniget.com/benchmark/omniget-vs-persepolis).

How do I make aria2 download faster?

Raise the connections per server and the split, for example -x16 -s16 -k1M; the default is one connection per server. In this test that saved about 3 seconds on a 372 MB file, inside the spread between rounds, because one connection already filled a line of about 37 Mbps. For torrents, adding more public trackers with --bt-tracker helped more.

How do I stop aria2 from seeding after a torrent finishes?

Add --seed-time=0. By default aria2 seeds until the share ratio reaches 1.0, so a plain aria2c magnet command does not exit at 100%. OmniGet 0.10.1 seeds until you remove the torrent from the list.

How do I install aria2 on a Mac?

Through Homebrew: brew install aria2. The project stopped publishing macOS builds after 1.35.0 in 2019. OmniGet has a macOS installer and does not need aria2 to download files or torrents.

## Other benchmarks

-   [OmniGet vs Videomass](https://getomniget.com/benchmark/omniget-vs-videomass)
-   [OmniGet vs ezytdl](https://getomniget.com/benchmark/omniget-vs-ezytdl)
-   [OmniGet vs Media Downloader](https://getomniget.com/benchmark/omniget-vs-media-downloader)
-   [OmniGet vs SCrawler](https://getomniget.com/benchmark/omniget-vs-scrawler)
-   [OmniGet vs Vividl](https://getomniget.com/benchmark/omniget-vs-vividl)
-   [OmniGet vs Hindsight](https://getomniget.com/benchmark/omniget-vs-hindsight)

## Try OmniGet 1.0.0

Free and open source, for Windows, macOS and Linux.

[Download](https://getomniget.com/download)
