picam
Features
- Generate H.264/AAC encoded MPEG-TS file from Raspberry Pi Camera (v1/v2) and optionally USB microphone or Wolfson Audio Card
- Generate HTTP Live Streaming files with optional encryption
- Add timestamp
- Display Unicode text with arbitrary font
Performance (Latency)
Time from real motion to playback on Strobe Media Player over RTMP:
Video bitrate | Minimum latency |
---|---|
300 Kbps | 0.3 sec |
500 Kbps | 0.6 sec |
1 Mbps | 0.8 sec |
2 Mbps | 1.0 sec |
3 Mbps | 1.3 sec |
In HTTP Live Streaming (HLS), the latency will never go below 3-4 seconds. This limitation stems from the design of HLS.
The above results were tested with:
- Video: 1280x720, 30 fps, GOP size 30
- Audio: 48 Khz mono, 40 Kbps
- RTMP Server: node-rtsp-rtmp-server
- Client: Flash Player 14,0,0,145 on Firefox 31.0 for Mac, using test/strobe_media_playback.html
- Network: Wi-Fi network created by a USB dongle attached to Raspberry Pi
Required hardware
- Raspberry Pi
- Raspberry Pi Camera Board v1 or v2
- (optionally) USB microphone or Wolfson Audio Card
Supported operating systems
- Raspbian
- Arch Linux
Installation
Binary release is available at https://github.com/iizukanao/picam/releases/latest
Also, out-of-the-box SD card image for live streaming (picam + Raspbian + live streaming server) is available at https://github.com/iizukanao/picam-streamer
If you want to build picam yourself, see INSTALL.md.
Using a binary release
The fastest way to use picam is to use a binary release. To set up and use it, run the following commands on your Raspberry Pi (Raspbian). It will set up picam in ~/picam/
.
# If you have not enabled camera, enable it with raspi-config then reboot
sudo raspi-config
# Install dependencies
sudo apt-get update
sudo apt-get install libharfbuzz0b libfontconfig1
# Create directories and symbolic links
cat > make_dirs.sh <<'EOF'
#!/bin/bash
DEST_DIR=~/picam
SHM_DIR=/run/shm
mkdir -p $SHM_DIR/rec
mkdir -p $SHM_DIR/hooks
mkdir -p $SHM_DIR/state
mkdir -p $DEST_DIR/archive
ln -sfn $DEST_DIR/archive $SHM_DIR/rec/archive
ln -sfn $SHM_DIR/rec $DEST_DIR/rec
ln -sfn $SHM_DIR/hooks $DEST_DIR/hooks
ln -sfn $SHM_DIR/state $DEST_DIR/state
EOF
chmod +x make_dirs.sh
./make_dirs.sh
# Optionally, increase microphone volume with alsamixer
alsamixer
# Install picam binary
wget https://github.com/iizukanao/picam/releases/download/v1.4.6/picam-1.4.6-binary.tar.xz
tar xvf picam-1.4.6-binary.tar.xz
cp picam-1.4.6-binary/picam ~/picam/
# Run picam
cd ~/picam
./picam --alsadev hw:1,0
Usage
Create symbolic links (optional, but strongly recommended)
You can take advantage of RAM drive (/run/shm/
) and reduce access to SD card. It also provides better quality of recording.
First, stop picam if it is running. Create rec, hooks, and state directories in /run/shm/
, then change directories with the same name in picam to symbolic links. Create another symbolic link from /run/shm/rec/archive
to somewhere on SD card.
Result:
picam
| ...
|-- archive
|-- hooks -> /run/shm/hooks
|-- rec -> /run/shm/rec
`-- state -> /run/shm/state
/run/shm/
|-- hooks
|-- rec
| |-- archive -> /home/pi/picam/archive
| `-- tmp (automatically created by picam)
`-- state
Finding ALSA device name
First, find ALSA device name of your microphone.
$ arecord -l
**** List of CAPTURE Hardware Devices ****
card 1: Device [USB PnP Sound Device], device 0: USB Audio [USB Audio]
Subdevices: 1/1
Subdevice #0: subdevice #0
ALSA device name is consisted of hw:<card>,<device>
. In the above example, the ALSA device name is hw:1,0
.
If you got no soundcards found
error, try sudo arecord -l
. If that output looks good, you might want to add your user to audio
group.
$ sudo usermod -a -G audio $USER
(once logout, then login)
$ groups
wheel audio pi <-- (make sure that 'audio' is in the list)
$ arecord -l
**** List of CAPTURE Hardware Devices ****
card 1: Device [USB PnP Sound Device], device 0: USB Audio [USB Audio]
Subdevices: 1/1
Subdevice #0: subdevice #0
Starting picam
Run picam with your ALSA device name.
$ ./picam --alsadev hw:1,0
configuring devices
capturing started
Recording
To start recording, create a file named hooks/start_record
while picam command is running.
$ touch hooks/start_record
You will see smth like disk_usage=23% start rec to rec/archive/2017-08-05_16-41-52.ts
in the picam command output.
To stop recording, create a file named hooks/stop_record
.
$ touch hooks/stop_record
You will see stop rec
in the picam command output.
The recorded MPEG-TS file is in rec/archive/
directory.
To convert MPEG-TS to MP4, run:
ffmpeg -i test.ts -c:v copy -c:a copy -bsf:a aac_adtstoasc test.mp4
# or
avconv -i test.ts -c:v copy -c:a copy -bsf:a aac_adtstoasc test.mp4
Mute/Unmute
To mute microphone temporarily, create a file named hooks/mute
.
$ touch hooks/mute
To unmute microphone, create a file named hooks/unmute
.
$ touch hooks/unmute
Command options
$ picam --help
picam version 1.4.6
Usage: picam [options]
Options:
[video]
-w, --width <num> Width in pixels (default: 1280)
-h, --height <num> Height in pixels (default: 720)
-v, --videobitrate <num> Video bit rate (default: 2000000)
Set 0 to disable rate control
-f, --fps <num> Frame rate (default: 30.0)
-g, --gopsize <num> GOP size (default: same value as fps)
--vfr Enable variable frame rate. GOP size will be
dynamically controlled.
--minfps <num> Minimum frames per second. Implies --vfr.
It might not work if width / height >= 1.45.
--maxfps <num> Maximum frames per second. Implies --vfr.
It might not work if width / height >= 1.45.
--rotation <num> Image rotation in clockwise degrees
(0, 90, 180, 270)
--hflip Flip image horizontally
--vflip Flip image vertically
--avcprofile <str> Set AVC/H.264 profile to one of:
constrained_baseline/baseline/main/high
(default: constrained_baseline)
--avclevel <value> Set AVC/H.264 level (default: 3.1)
--qpmin <num> Minimum quantization level (0..51)
--qpmax <num> Maximum quantization level (0..51)
--qpinit <num> Initial quantization level
--dquant <num> Slice DQuant level
[audio]
-c, --channels <num> Audio channels (1=mono, 2=stereo)
Default is mono. If it fails, stereo is used.
-r, --samplerate <num> Audio sample rate (default: 48000)
-a, --audiobitrate <num> Audio bit rate (default: 40000)
--alsadev <dev> ALSA microphone device (default: hw:0,0)
--volume <num> Amplify audio by multiplying the volume by <num>
(default: 1.0)
--noaudio Disable audio capturing
--audiopreview Enable audio preview
--audiopreviewdev <dev> Audio preview output device (default: plughw:0,0)
[HTTP Live Streaming (HLS)]
-o, --hlsdir <dir> Generate HTTP Live Streaming files in <dir>
--hlsenc Enable HLS encryption
--hlsenckeyuri <uri> Set HLS encryption key URI (default: stream.key)
--hlsenckey <hex> Set HLS encryption key in hex string
(default: 75b0a81de17487c88a47507a7e1fdf73)
--hlsenciv <hex> Set HLS encryption IV in hex string
(default: 000102030405060708090a0b0c0d0e0f)
[output for node-rtsp-rtmp-server]
--rtspout Enable output for node-rtsp-rtmp-server
--rtspvideocontrol <path> Set video control socket path
(default: /tmp/node_rtsp_rtmp_videoControl)
--rtspaudiocontrol <path> Set audio control socket path
(default: /tmp/node_rtsp_rtmp_audioControl)
--rtspvideodata <path> Set video data socket path
(default: /tmp/node_rtsp_rtmp_videoData)
--rtspaudiodata <path> Set audio data socket path
(default: /tmp/node_rtsp_rtmp_audioData)
[MPEG-TS output via TCP]
--tcpout <url> Enable TCP output to <url>
(e.g. --tcpout tcp://127.0.0.1:8181)
[camera]
--autoex Enable automatic control of camera exposure between
daylight and night modes. This forces --vfr enabled.
--autoexthreshold <num> When average value of Y (brightness) for
10 milliseconds of captured image falls below <num>,
camera exposure will change to night mode. Otherwise
camera exposure is in daylight mode. Implies --autoex.
(default: 5.0)
If --verbose option is enabled as well, average value of
Y is printed like y=28.0.
--ex <value> Set camera exposure. Implies --vfr. <value> is one of:
off auto night nightpreview backlight spotlight sports
snow beach verylong fixedfps antishake fireworks
largeaperture smallaperture
--wb <value> Set white balance. <value> is one of:
off: Disable white balance control
auto: Automatic white balance control (default)
sun: The sun provides the light source
cloudy: The sun provides the light source through clouds
shade: Light source is the sun and scene is in the shade
tungsten: Light source is tungsten
fluorescent: Light source is fluorescent
incandescent: Light source is incandescent
flash: Light source is a flash
horizon: Light source is the sun on the horizon
--wbred <num> Red gain. Implies "--wb off". (0.0 .. 8.0)
--wbblue <num> Blue gain. Implies "--wb off". (0.0 .. 8.0)
--metering <value> Set metering type. <value> is one of:
average: Center weight average metering (default)
spot: Spot (partial) metering
matrix: Matrix or evaluative metering
backlit: Assume a backlit image
--evcomp <num> Set Exposure Value compensation (-24..24) (default: 0)
--shutter <num> Set shutter speed in microseconds (default: auto).
Implies --vfr.
--iso <num> Set ISO sensitivity (100..800) (default: auto)
--roi <x,y,w,h> Set region of interest (crop rect) in ratio (0.0-1.0)
(default: 0,0,1,1)
-p, --preview Display fullscreen preview
--previewrect <x,y,width,height>
Display preview window at specified position
--opacity Preview window opacity
(0=transparent..255=opaque; default=255)
--blank[=0xAARRGGBB] Set the video background color to black (or optional ARGB value)
--query Query camera capabilities then exit
[timestamp] (may be a bit heavy on Raspberry Pi 1)
--time Enable timestamp
--timeformat <spec> Timestamp format (see "man strftime" for spec)
(default: "%a %b %d %l:%M:%S %p")
--timelayout <spec> Timestamp position (relative mode)
layout is comma-separated list of:
top middle bottom left center right
(default: bottom,right)
--timehorizmargin <px> Horizontal margin from edge (default: 10).
Effective only if --timelayout is used.
--timevertmargin <px> Vertical margin from edge (default: 10).
Effective only if --timelayout is used.
--timepos <x,y> Timestamp position (absolute mode)
--timefontname <name> Timestamp font name (default: FreeMono:style=Bold)
--timefontfile <file> Timestamp font file. This invalidates --timefontname.
--timefontface <num> Timestamp font face index (default: 0).
Effective only if --timefontfile is used.
--timept <pt> Text size in points (default: 14.0)
--timedpi <num> DPI for calculating text size (default: 96)
--timecolor <hex> Text color (default: ffffff)
--timestrokecolor <hex> Text stroke color (default: 000000)
Note that texts are rendered in grayscale.
--timestrokewidth <pt> Text stroke border radius (default: 1.3).
To disable stroking borders, set this value to 0.
--timespacing <px> Additional letter spacing (default: 0)
[misc]
--recordbuf <num> Start recording from <num> keyframes ago
(must be >= 1; default: 5)
--statedir <dir> Set state dir (default: state)
--hooksdir <dir> Set hooks dir (default: hooks)
-q, --quiet Suppress all output except errors
--verbose Enable verbose output
--version Print program version
--help Print this help
White balance
Camera white balance can be set either via command line option (e.g. --wb sun
) or hooks. To change the white balance while picam is running, create hooks/wb_<value>
, where <value>
is the name of white balance.
For example, the following command will dynamically change the white balance to sun.
$ touch hooks/wb_sun
Available white balance modes are: off, auto, sun, cloudy, shade, tungsten, fluorescent, incandescent, flash, and horizon.
Exposure Control
Camera exposure control can be set either via command line option (e.g. --ex night
) or hooks. To change the exposure control while picam is running, start picam with --vfr
or --ex
option, then create hooks/ex_<value>
, where <value>
is the name of exposure control.
For example, the following command will dynamically change the exposure control to night.
$ touch hooks/ex_night
For the list of available exposure control values, see picam --help
.
Value | Description |
---|---|
off |
Disable exposure control |
auto |
Automatic exposure |
night |
Exposure at night |
nightpreview |
Shorter exposure than night |
backlight |
Exposure with backlight illuminating the subject |
spotlight |
Exposure with a spotlight illuminating the subject |
sports |
Exposure for sports |
snow |
Exposure for the subject in snow |
beach |
Exposure for the subject at a beach |
verylong |
Long exposure |
fixedfps |
Constrain FPS to a fixed value |
antishake |
Antishake mode |
fireworks |
Optimized for fireworks |
largeaperture |
Exposure when using a large aperture on the camera |
smallaperture |
Exposure when using a small aperture on the camera |
Recordbuf
recordbuf
is a parameter which controls how many past keyframes should be included at the start of a recording. For example, recordbuf=1
means that a recording will start from the last keyframe, and recordbuf=2
means that a recording will start from the second last keyframe relative to when hooks/start_record
is created. The minimum value of recordbuf
is 1.
Global and per-recording recordbuf
Added in version 1.4.0
There are two types of recordbuf; global and per-recording. Global recordbuf is the default value for all recordings. Per-recording recordbuf only applies to the current recording. Per-recording recordbuf must be less than or equal to global recordbuf.
Setting global recordbuf
Global recordbuf can be specified by either --recordbuf
option or hooks/set_recordbuf.
# Set global recordbuf to 30
echo 30 > hooks/set_recordbuf
Setting per-recording recordbuf
Per-recording recordbuf has a default value which is the same value as global recordbuf. Per-recording recordbuf can be specified via hooks/start_record
.
# Start recording with per-recording recordbuf set to 2
$ echo recordbuf=2 > hooks/start_record
Overlaying text (subtitle)
Added in version 1.4.0
picam can display text with correct ligatures and kerning, with a font of your choice. To display a text, create hooks/subtitle.
$ echo 'text=Houston, we have a problem' > hooks/subtitle
Each line of hooks/subtitle must be in the format of key=value
. Lines starting with #
will be ignored. Supported keys are:
Key | Description | Default value |
---|---|---|
text | UTF-8 encoded text. \n will be treated as a line break. |
(none) |
font_name | Font name which can be recognized by Fontconfig | sans |
font_file | Path to font file. If this is specified, font_name will not be used. | (none) |
face_index | Font face index. Effective only if font_file is specified. | 0 |
pt | Text size in points | 28.0 |
dpi | DPI for calculating text size | 96 |
duration | Number of seconds the text appears on the screen. If this is set to 0, the text will be displayed indefinitely. | 7.0 |
layout_align | Layout alignment of the text box on the screen. Comma-separated list of: top middle bottom left center right | bottom,center |
horizontal_margin | Horizontal margin from the nearest edge in pixels. Does nothing when pos is specified or layout_align has "center". | 0 |
vertical_margin | Vertical margin from the nearest edge in pixels. Does nothing when pos is specified or layout_align has "middle". | 35 |
pos | Absolute position of the text box on the screen. This invalidates layout_align settings. | (none) |
text_align | Text alignment inside the positioned box (left, center, right) | center |
line_height | Line spacing is multiplied by this number | 1.0 |
tab_scale | Tab width is multiplied by this number | 1.0 |
letter_spacing | Additional letter spacing in pixels | 0 |
color | Text color in hex color code. Rendered in grayscale. | ffffff |
stroke_color | Text stroke (outline) color in hex color code. Rendered in grayscale. | 000000 |
stroke_width | Text stroke (outline) border radius in points | 1.0 |
in_preview | Visibility of the text in the preview | 1 |
in_video | Visibility of the text in the encoded video | 1 |
NOTE: On the first generation models of Raspberry Pi (before Pi 2), subtitles cause CPU usage high and the video frame rate may drop below 30 fps.
Examples
$ cat example1
text=What goes up\nmust come down\nfinally floor AV Wa
font_name=serif
pt=40
$ cat example1 > hooks/subtitle
$ cat example2
text=お気の毒ですが\n冒険の書は\n消えちゃいました☆
font_file=/home/pi/uzura.ttf
pt=46
$ cat example2 > hooks/subtitle
$ cat example3
text=♨☀♻♥⚠
font_file=/home/pi/NotoSansCJKjp-Regular.otf
pt=120
layout_align=middle,center
letter_spacing=40
$ cat example3 > hooks/subtitle
$ cat example4
text=●REC
font_name=FreeSans
pt=40
layout_align=top,right
horizontal_margin=30
vertical_margin=30
color=000000
stroke_width=0
duration=0
$ cat example4 > hooks/subtitle
Changing the filename for recording
Added in version 1.4.0
To change the directory and/or filename for the recorded file, specify dir
and/or filename
parameters in hooks/start_record
.
# Start recording to /tmp/myout.ts
$ echo -e "dir=/tmp\nfilename=myout.ts" > hooks/start_record
Determine the length of a recorded file
Added in version 1.4.0
The file state/recorded_filename has some info about the recording.
$ cat state/2015-11-19_01-18-09.ts
duration_pts=2083530
duration_sec=23.150333
You can remove state/*.ts
files if you do not need them.
HTTP Live Streaming (HLS)
HTTP Live Streaming is disabled by default. To enable HTTP Live Streaming and generate files in /run/shm/hls, run:
$ ./picam -o /run/shm/hls
Serving HLS
Set up nginx (ignore "Additional - Install PHP" step), then open /etc/nginx/sites-available/default with a text editor and add the following code inside server { ... }
block.
location /hls/ {
root /run/shm;
}
Restart the nginx server with sudo service nginx restart
then run picam with -o /run/shm/hls
option. The HLS will be available at http://YOUR-PI-IP/hls/index.m3u8
Enabling encryption
Optionally you can enable encryption for HTTP Live Streaming. We will use the following settings as an example.
- HTTP Live Streaming output directory:
/run/shm/hls/
- Encryption key:
0xf0f1f2f3f4f5f6f7f8f9fafbfcfdfeff
- Encryption IV:
0x000102030405060708090a0b0c0d0e0f
- Encryption key file:
enc.key
First you have to create a file named enc.key
which contains 16-byte encryption key. To create such file, run:
$ echo -n $'\xf0\xf1\xf2\xf3\xf4\xf5\xf6\xf7\xf8\xf9\xfa\xfb\xfc\xfd\xfe\xff' > enc.key
Put enc.key
in /run/shm/hls/
directory. Then, run picam with the following options:
$ ./picam -o /run/shm/hls --hlsenc --hlsenckeyuri enc.key \
--hlsenckey f0f1f2f3f4f5f6f7f8f9fafbfcfdfeff \
--hlsenciv 000102030405060708090a0b0c0d0e0f
You can watch the HTTP Live Streaming by accessing /run/shm/hls/index.m3u8
via HTTP or HTTPS with QuickTime Player.
Using picam in combination with nginx-rtmp-module
To use picam with nginx-rtmp-module, add the following lines to nginx.conf
:
rtmp {
server {
listen 1935;
chunk_size 4000;
application webcam {
live on;
exec_static /path/to/ffmpeg -i tcp://127.0.0.1:8181?listen
-c:v copy -ar 44100 -ab 40000
-f flv rtmp://localhost:1935/webcam/mystream;
}
}
}
Note that /path/to/ffmpeg
should be replaced with the actual absolute path to ffmpeg command.
Start nginx server, then run:
$ ./picam --tcpout tcp://127.0.0.1:8181
You can access your live stream at rtmp://YOUR_RASPBERRYPI_IP/webcam/mystream
.
Publishing to Ustream
To upload streams from picam to Ustream, run ffmpeg with the following options. RTMP_URL
and STREAM_KEY
can be obtained from Ustream's Channel settings → Broadcast settings → Encoder settings.
$ ffmpeg -i tcp://127.0.0.1:8181?listen -c:v copy -c:a aac -strict -2 -ar 44100 -ab 40000 -f flv RTMP_URL/STREAM_KEY
Then, run picam to start streaming.
$ picam --tcpout tcp://127.0.0.1:8181
Recommended hardware
USB microphone
Any cheap USB microphone should work as long as it is supported by ALSA. I have tested this program with the combination of:
- USB to 3.5mm audio adapter: PLANEX PL-US35AP
- Microphone: ELECOM MS-STM95
License
LGPL v2.1, as some parts of FFmpeg source is included.