Hardware & transcoding
How hard does your server work?
Most playback is direct, which costs the server almost nothing. Transcoding happens only when a device needs something the file doesn't already have. This page explains what that means for the hardware you run.
Direct play vs transcoding
Direct play, the ideal path
When your client can already decode the file, Quven streams it untouched: full quality, instant start, and almost no work for the server.
Transcoding, only when needed
If a device needs a different codec, container or bitrate, the server converts the stream on the fly. That is the work a faster CPU or a GPU speeds up.
Subtitle recognition, occasional and heavy
Writing a subtitle from the dialogue is the heaviest thing the server does, but it runs once per file and only when you ask. A GPU turns a long wait into a short one.
What you need to run the server
- OS
- Windows 10 1809+ / macOS 13+ / Linux with glibc 2.34 or newer and OpenSSL 3, so Ubuntu 22.04+, Debian 12+, Fedora, or RHEL, AlmaLinux and Rocky 9+
- CPU
- Windows: x86-64 or ARM64; Linux: x86-64 or ARM64; macOS: Intel (x64) or Apple silicon (ARM64). A GPU accelerates transcoding and subtitle speech recognition
- RAM
- 4 GB minimum, 8 GB recommended
- Disk
- ~1 GB for the app + room for metadata & artwork; the optional speech model adds about 0.5 GB when first requested
- Runtime
- Everything included, nothing else to install
- Network
- A LAN for local play; internet for first sign-in, token refresh, metadata and cloud features
Getting the most from your hardware
- Native clients direct-play the widest range of formats, so they transcode the least.
- A dedicated GPU accelerates transcoding and HDR tone-mapping; the CPU path is always there as a universal fallback.
- Every simultaneous transcode wants more CPU or GPU headroom, and direct play scales much further on the same box.
- Keep your library on the same machine or a fast network share so scanning and playback stay snappy.
- Speech recognition uses Metal on macOS and Vulkan on Windows and Linux where a usable adapter exists, and always leaves one processor to the rest of the system.
Bring your library home.
Free for home use and yours to run on your own hardware. Install it in minutes.