Better and Cheaper Than IPTV
Better and Cheaper Than IPTV
Imagine buying a very nice TV and figuring out it doesn't allow casting from random websites... Then you switch to the longest HDMI cable you can find. Then you start doing screen mirroring: your computer lags, resolution tanks, nothing feels right.
Castor is a CLI that extracts video streams from websites, handles format compatibility, and casts to your TV in real time, with optional auto-generated subtitles burned directly into the video. Run castor cast to browse trending titles, search TMDB, inspect posters and metadata, then cast - without leaving the terminal.
How extraction works
Castor launches a headless Chrome with a randomized browser fingerprint and stealth scripts to hide automation. It listens to all network activity via Chrome DevTools Protocol, captures video streams, then runs an action pipeline: click the page, navigate into the largest iframe, solve Cloudflare Turnstile if detected, and click again as fallback. This works on most streaming websites but won't beat sophisticated bot protection.
Installation
Homebrew and source build a native binary that needs Chrome/Chromium (headless extraction), ffmpeg (transcoding), and ffprobe (format detection) on your PATH. The Docker image bundles all three, so there you only supply a config.
brew install --cask stupside/tap/castor
ghcr.io/stupside/castor
Ships with Chrome, ffmpeg and ffprobe baked in.
Device discovery (no config required)
docker run --rm --network host ghcr.io/stupside/castor:latest scan
Cast a movie by id
Mounting config.yaml and a persistent model cache:
docker run --rm --network host \
-v "$PWD/config.yaml:/config.yaml" \
-v castor-cache:/root/.cache \
ghcr.io/stupside/castor:v1.4.0 \
cast movie tt12300742
The -v "$PWD/config.yaml:/config.yaml" mount is what makes this work: Castor reads your device and sources from config.yaml, which the container looks for at /config.yaml. cast movie tt12300742 builds the player URL from the sources proxies in that file - no URL on the command line - so run every command from the directory holding your config.yaml.
Warning
Streaming sites are volatile. cast movie resolves the id against the sources proxies in your config.yaml - here https://1embed.cc. Those point at third-party streaming sites that can go offline, change domains, or start blocking at any time. When one stops resolving, rotate it: swap in a working mirror/domain in the sources.proxies list in your config.yaml.
Important
--network host is required: device discovery is SSDP multicast and the TV streams back from Castor's replay server - neither survives Docker's bridge network. Host networking is only real on Linux; on Docker Desktop (macOS/Windows) it won't reach your TV, so run the binary natively there instead.
The castor-cache volume keeps the auto-downloaded whisper models (~75 MB) between runs. Swap :v1.4.0 for :latest to track the newest build, or for any other tag to pin a release.
Building from source
Needs Go 1.26+ and cmake (the whisper.cpp bindings are cgo and link a locally built libwhisper.a):
git clone --recurse-submodules https://github.com/stupside/castor.git
cd castor
make # builds libwhisper.a, then the castor binary
go install won't work: the vendored whisper.cpp bindings come in through a local replace and need that prebuilt static lib.
Compatible devices
Any TV implementing the DLNA/UPnP MediaRenderer:1 profile works, which covers virtually every smart TV sold in the last decade: Samsung (tested), LG, Sony Bravia, Panasonic Viera, Philips, Hisense, TCL, VIZIO, Sharp. Networked players like Kodi, VLC, and Plex also work. Run castor scan to discover devices on your network.
Experimental
Implemented but untested. Contributions welcome.
Configuration
Castor requires a config.yaml in the current directory (or pass --config). Everything mechanical ships with working defaults, so the file only has to say which device to cast to, which sources to cast from, and a free TMDB API key for the browser.
Find your TV's exact name:
castor scanCreate
config.yamlwith that name:device: name: "Living Room TV" # exact name from `castor scan` type: dlna sources: - proxies: ["https://vidsrc-embed.ru"] templates: movie: "/embed/movie/{itemID}" episode: "/embed/tv/{itemID}/{season}-{episode}" tmdb: api_key: " " # free from https://www.themoviedb.org/settings/apiBrowse and cast from an interactive TUI:
castor cast
castor cast first asks which device to cast to - every DLNA/UPnP renderer on your network, discovered on the fly and with your configured device pre-selected. Then it opens a TMDB-backed browser - filter by genre, search, inspect posters and metadata, drill into a series' episodes, and cast the one you pick.
Tip
No TMDB key? You can skip the browser and cast a title straight from its id - castor cast movie tt33028778 - using the sources in your config. See Usage.
Usage examples
# Interactive TMDB browser: search, pick a movie/episode, cast (needs tmdb.api_key)
castor cast
# Cast from a streaming site's player page
castor cast player https://1embed.cc/embed/movie/tt12300742
# Cast by IMDB/TMDB id, using the sources in your config
castor cast movie tt33028778
castor cast episode tt2699128 --season 1 --episode 3
# Cast a raw stream URL directly
castor cast url https://example.com/stream.m3u8
# Useful flags
castor cast movie --dry-run tt33028778 # print found URLs without casting
castor --debug cast player https://... # verbose logging
castor scan # discover devices on the network
castor info # version / build info
Configuration details
Quick start covers the required keys. Beyond those, everything mechanical (timeouts, probing, capture, transcoding, Chrome discovery) ships with working defaults. Override any of it in config.yaml, point at a different file with --config, drop secrets like your TMDB key into a git-ignored sibling config.local.yaml (it overlays config.yaml), or set CASTOR_SECTION__FIELD environment variables.
The one opt-in worth calling out is auto-generated subtitles, burned into the video:
whisper:
enable: true # off by default
# language: "fr" # default: English
# model_path: "" # default: ggml-tiny.en (~75 MB, auto-downloaded)
See CONTRIBUTING.md.
Comments
No comments yet. Start the discussion.