- Proper sound support: Syncing of multiple unrelated sources through
high-quality resampling, freely selectable input, cue out for headphones,
- dynamic range compression, simple EQ (lowpass), level meters conforming
+ dynamic range compression, simple EQ (low-cut), level meters conforming
to EBU R128.
- Theme engine encapsulating the design demands of each individual
- Two or more Blackmagic USB3 or PCI cards, either HDMI or SDI.
The PCI cards need Blackmagic's own drivers installed. The USB3 cards
are driven through the “bmusb” driver embedded in bmusb/, using libusb-1.0.
- Note that you will want a recent Linux kernel to avoid LPM (link power
- management) and bandwidth allocation issues with USB3.
+ You want a recent kernel and libusb-1.0; see below.
- Movit, my GPU-based video filter library (https://movit.sesse.net).
You will need at least version 1.3.1.
- libmicrohttpd for the embedded web server.
+ - x264 for encoding high-quality video suitable for streaming to end users.
+
- ffmpeg for muxing, and for encoding audio.
- Working OpenGL; Movit works with almost any modern OpenGL implementation.
- Nageru has been tested with Intel on Mesa 10.x and 11.x (you probably want
- 11.x), and with NVIDIA's proprietary drivers. AMD's proprietary drivers
- (fglrx) are known not to work due to driver bugs; I am in contact with
- AMD to try to get this resolved.
+ Nageru has been tested with Intel on Mesa (you want 11.2 or newer, due
+ to critical stability bugfixes), and with NVIDIA's proprietary drivers.
+ AMD's proprietary drivers (fglrx) are known not to work due to driver bugs;
+ I am in contact with AMD to try to get this resolved.
- libzita-resampler, for resampling sound sources so that they are in sync
between sources, and also for oversampling for the peak meter.
git submodule update --init
apt install qtbase5-dev qt5-default pkg-config libmicrohttpd-dev \
libusb-1.0-0-dev liblua5.2-dev libzita-resampler-dev libva-dev \
- libavcodec-dev libavformat-dev libswscale-dev libmovit-dev \
- libegl1-mesa-dev libasound2-dev
+ libavcodec-dev libavformat-dev libswscale-dev libavresample-dev \
+ libmovit-dev libegl1-mesa-dev libasound2-dev libx264-dev
+
+Exceptions as of June 2016:
+
+ - You need libusb 1.0.21 or newer (not yet released; you will need to
+ use git) if you want zerocopy USB. Zerocopy USB helps not only for
+ performance, but also for stability.
+
+ - Nageru depends on an avformat API for marking block boundaries in the
+ muxed byte stream that didn't enter ffmpeg before
+ 4e7a9212820a56bc731c09b2f11ae1422d070837, pushed to git late June 2016.
-The patches/ directory contains some patches for upstream software that help
-Nageru performance and/or stability. They are all meant for upstream, but
-probably will not be in by the time Nageru is released. All except the bmusb
-patch are taken to be by Steinar H. Gunderson <sesse@google.com> (ie., my work
-email, unlike Nageru itself and bmusb), and under the same license as the
-projects they patch.
+The patches/ directory contains a patch that helps zita-resampler performance.
+It is meant for upstream, but was not in at the time Nageru was released.
+It is taken to be by Steinar H. Gunderson <sesse@google.com> (ie., my ex-work
+email), and under the same license as zita-resampler itself.
To start it, just hook up your equipment, type “make” and then “./nageru”.
It is strongly recommended to have the rights to run at real-time priority;
performance. The same goes for PulseAudio.
Nageru will open a HTTP server at port 9095, where you can extract a live
-H264+PCM signal in QuickTime mux (e.g. http://127.0.0.1:9095/stream.mov).
+H264+PCM signal in nut mux (e.g. http://127.0.0.1:9095/stream.nut).
It is probably too high bitrate (~25 Mbit/sec depending on content) to send to
users, but you can easily send it around in your internal network and then
transcode it in e.g. VLC. A copy of the stream (separately muxed) will also
be saved live to local disk.
+If you have a fast CPU (typically a quadcore desktop; most laptops will spend
+most of their CPU on running Nageru itself), you can use x264 for the outgoing
+stream instead of Quick Sync; it is much better quality for the same bitrate,
+and also has proper bitrate controls. Simple add --http-x264-video on the
+command line. (You may also need to add something like "--x264-preset veryfast",
+since the default "medium" preset might be too CPU-intensive, but YMMV.)
+The stream saved to disk will still be the Quick Sync-encoded stream, as it is
+typically higher bitrate and thus also higher quality. Note that if you add
+".metacube" at the end of the URL (e.g. "http://127.0.0.1:9095/stream.ts.metacube"),
+you will get a stream suitable for streaming through the Cubemap video reflector
+(cubemap.sesse.net). A typical example would be:
+
+ ./nageru --http-x264-video --x264-preset veryfast --x264-tune film \
+ --http-mux mp4 --http-audio-codec libfdk_aac --http-audio-bitrate 128
+
+If you are comfortable with using all your remaining CPU power on the machine
+for x264, try --x264-speedcontrol, which will try to adjust the preset
+dynamically for maximum quality, at the expense of somewhat higher delay.
+
+See --help for more information on options in general.
The name “Nageru” is a play on the Japanese verb 投げる (nageru), which means
to throw or cast. (I also later learned that it could mean to face defeat or
projects / nageru / blobdiff