Riptide

A free, open source download manager.

Multi-connection downloads, resume that cannot corrupt your file, browser capture, and HLS/DASH video — with no licence key, no trial, and no kernel driver.

---

What it does

• Dynamic multi-connection downloads. A file is not cut into fixed parts. Riptide starts with one range and, each time a connection frees up, carves a new segment out of whichever range has the most work left. A connection that slows down accumulates outstanding bytes and becomes the next thing to split, so one bad route cannot hold the whole download back.
• No reassembly pass. Every segment writes straight into its final position in one sparse file. When the last byte lands the download is finished — no temp directory, no concatenation step, no sitting at 100% while gigabytes are copied around.
• Resume that is actually safe. Progress is checkpointed beside the file. Before reusing a partial download Riptide re-probes the URL and compares the ETag, Last-Modified and size, and refuses if the remote file changed rather than splicing new bytes into stale ones.
• Browser capture. A Manifest V3 extension hands downloads from Chrome, Edge or Brave to Riptide, carrying the referer and — if you allow it — the site's cookies, so links from logged-in pages still work.
• HLS and DASH. Adaptive streams are parsed, quality variants listed, and segments fetched with look-ahead into a single file. Plain AES-128 is decrypted; DRM is detected and refused.
• Queue, speed limits, categories, mirrors, proxies. A local web UI to drive it all, and a CLI that works entirely on its own.

No telemetry. Riptide makes no network request you did not ask for.

Quick start

git clone https://github.com/ZekkJJ/riptide
cd riptide
cargo build --release

Two binaries land in target/release/:

./target/release/riptide https://example.com/big-file.iso

./target/release/riptided --open

riptide is the command line tool. riptided is the background service: it runs the queue, serves a web UI at http://127.0.0.1:7739/, and is what the browser extension talks to.

Rust 1.89 or newer is the only requirement. TLS is rustls and the crypto is pure Rust, so there is no OpenSSL to install.

Command line

riptide <url>                                   # download, 8 connections by default
riptide <url> -n 16 -o name.iso -d ~/Downloads  # connections, name, folder
riptide <url> -l 2M                             # cap at 2 MiB/s
riptide <url> --categorize                      # file into Video/, Audio/, Archives/, ...
riptide resume ./partly-downloaded.iso          # pick up where it stopped
riptide variants https://site/stream.m3u8       # list the quality levels
riptide https://site/stream.m3u8 --variant 1080 # fetch a stream as one file
riptide <url> --referer <page> --cookies "s=…"  # links behind a login
riptide <url> --mirror <url2> --mirror <url3>   # fall back between mirrors
riptide <url> --proxy socks5://127.0.0.1:1080

Interrupting with Ctrl-C leaves the .rtpart and .rtstate files in place; riptide resume continues from there.

Browser extension

1. Start the service: riptided. It prints a connection token.
2. Open chrome://extensions, turn on Developer mode, choose Load unpacked, and select the extension/ folder.
3. Open the extension's options, paste the token, and press Save and test.

Downloads the browser starts are then handed to Riptide automatically, and right-click → Download with Riptide always works on any link.

Cookie capture for logged-in sites is a separate, optional permission you grant from the options page. Without it Riptide still downloads; some sites will simply refuse links that need a session.

A note on the local port

The extension reaches Riptide over http://127.0.0.1:7739, which means any page in your browser can send requests to it too. Two things stop that mattering:

• Every API route requires the X-Riptide-Token header. The token is generated per install, lives only in your config file, is compared in constant time, and is never served over HTTP.
• The web UI has the token inlined and is served with no CORS headers, so a cross-origin fetch of it is blocked by the browser from reading the response.

The listener binds 127.0.0.1 explicitly, so nothing is reachable from your network. If you would rather not have a local port at all, use the CLI — it talks to no one.

Configuration

riptided config-path prints the location (%APPDATA%\riptide\config.json on Windows, ~/.config/riptide/config.json elsewhere). Everything is editable from the web UI's Settings dialog, or directly:

Setting	Meaning
download_dir	Where finished files go
categorize	File into Video/, Audio/, Archives/, …
connections_per_download	Segments per file (default 8)
max_concurrent_downloads	Files transferring at once (default 3)
global_speed_limit	Bytes per second across everything, 0 for unlimited
max_retries	Attempts per segment, and for the initial probe
proxy	http://host:8080 or socks5://host:1080
capture_min_size	Smallest response worth taking over from the browser
capture_exclusions	Host globs never captured
categories	Extension → folder mapping

Why eight connections, and not eighty

More connections stop helping once the link is saturated, and many servers rate-limit or blacklist clients that open a lot of them at once — especially across several simultaneous downloads. Eight is a good default. Riptide will not split a range below 1 MiB, so small files quietly use fewer.

If a server will not serve byte ranges, or will not state a length, Riptide detects that during the probe and uses a single connection rather than producing a corrupt file.

How it is put together

crates/riptide-core/     the engine
  chunks.rs              dynamic work-stealing segmentation  (the heart)
  probe.rs               size, range support, filename, resume validators
  download.rs            connection pool, retries, mirrors, pause/resume
  storage.rs             sparse file, positional writes, no reassembly
  state.rs               resume state and the "did the file change?" check
  media.rs               HLS and DASH manifest parsing
  stream.rs              segment fetching, AES-128, stream resume
  category.rs            extension → folder
  throttle.rs            token bucket and speed measurement
crates/riptide-cli/      the `riptide` command
crates/riptide-daemon/   the `riptided` service, HTTP API and queue
extension/               Manifest V3 browser extension
web/                     the local web UI
docs/ANALYSIS.md         design rationale: why the engine is shaped this way

cargo test --workspace

126 tests, covering the chunk-tiling invariants, out-of-order segment reassembly, resume validation, filename sanitisation (including traversal attempts in server-supplied names), manifest parsing, AES-128 round trips, rate limiting under a paused clock, and API authentication. No test touches the network.

Verified end to end: an eight-connection download is SHA-256 identical to the same file fetched over one connection; a download killed at 18% resumes to the same hash; and a 64-segment 1080p HLS stream produces a valid 502 MB transport stream with correct packet sync at both ends.

What it does not do yet

Being straight about the gaps:

• FTP — the engine is HTTP/HTTPS only so far.
• Automatic link recapture. When a signed URL expires Riptide says so clearly, but does not yet ask the browser to go and capture a fresh one.
• Site grabber — no crawler for pulling every linked file off a page.
• ZIP preview — fetching the central directory first to show a ZIP's contents mid-download is a good trick and is not implemented.
• Media conversion — no transcoding. Stream segments are concatenated, which gives a playable file for fMP4 and MPEG-TS but is not a remux.
• Live streams are detected and refused rather than recorded.
• Scheduler — no "start at 2am" yet; the queue is manual.
• Firefox — the extension is Chromium-only for now. The APIs it uses are standard, so a port is mostly manifest work.

Contributing

See CONTRIBUTING.md. Briefly: sign your commits off (DCO), keep cargo test, cargo clippy and cargo fmt clean, and never contribute code derived from a proprietary program.

A few things are settled decisions and will not be merged: DRM circumvention, anything touching another program's licensing, kernel drivers, broader extension permissions, and telemetry.

Licence and legal

Riptide is licensed under the GNU General Public License, version 3 or later — see LICENSE. Dependency licences are inventoried in THIRD-PARTY-LICENSES.md and enforced in CI by cargo deny.

Riptide is an independent project, not affiliated with, endorsed by, or derived from Tonec FZE or Internet Download Manager. It contains no code, resources or data from any proprietary download manager. docs/ANALYSIS.md is a design-rationale document recording why the engine is built the way it is; it describes behaviour, not anyone's source code.

Riptide does not circumvent DRM — Widevine, PlayReady and FairPlay are detected and refused — and does not interact with any other software's licensing or activation.

Riptide is a general-purpose file transfer tool, like curl, wget or a web browser. You are responsible for what you download, and for respecting the law and the terms of the sites you download from.

Full notices, including trademarks and cryptography export, are in NOTICE.md.