Version 2.4.0

• Support for PulseAudio output via new output type pulse. With this
  feature you can eg. play the sound via the soundcard of the Raspberry
  Pi you run RTLSDR-Airband on (you need to install and run pulseaudio
  daemon on it, though). Or you can stream the audio from a Pi located
  near the antenna (eg. in the attic) to speakers connected to the desktop
  PC you are sitting at, without launching a local Icecast server,
  as before. Because the audio stream is sent uncompressed, it is
  not recommended to run it across the Internet - jitter or packet loss
  will easily cause the audio to become choppy. However in a local network
  PulseAudio is a good choice. And it gives much lower latency as compared
  to Icecast (typically under 0.5 seconds). Thanks to Marcus Ströbel
  for the idea and initial implementation.
• Support for referring to RTL devices by their serial numbers in the
  config file. Instead of "index = <dongle_index>" parameter, use "serial =
  <dongle_serial_number>" to get consistent behavior across reboots
  and hardware reconfigurations.
• Set RTL gain to the nearest gain value supported by the device. This is
  required for E4000 tuners, which do not round the given gain value to
  the nearest supported setting, which causes the gain setting operation
  to fail.
• Improved squelch operation in scan mode. All squelch-related variables
  (noise floor, AGC coefficients, etc) are now calculated and stored
  separately for each scanned channel. Earlier their values were common
  to all channels, which caused squelch problems in case when noise floor
  varied considerably between channels. Thanks to @strix-technica.
• Added build target for FreeBSD on x86. Use PLATFORM=x86-freebsd to
  compile and PLATFORM=x86-freebsd gmake install to install. Thanks
  to @nyammy.
• Display squelch setting in waterfall in place of noise floor value when
  squelch is set manually.
• Bug fixes, performance improvements.
• Decluttered and more understandable documentation.

Version 2.3.0

• Added support for mixers. It is now possible to produce audio streams
  combined from several input channels. Both mono and stereo mixing is
  supported. Usage example is provided in config/mixers.conf. All
  mixer-related parameters are documented in config/reference.conf.
• Added build options for 64-bit ARM architectures, like Odroid C2.
  Please use PLATFORM=armv8-generic when compiling.
• Fixed a long-standing bug in RTL sample processing, which caused some
  samples to be processed twice. If you were annoyed by these regular
  clicks in NFM audio every 125 ms, they are now gone.
• Reduced CPU usage on x86
• Some code restructuring and cleanups
• Added several configuration file examples for typical real-life
  scenarios. They are placed in config/ subdirectory. rtl_airband.conf.example
  file has been moved to config/reference.conf. It is meant to be a reference
  for all supported config knobs together with their description. This is
  still an interim solution before some more readable and understandable
  documentation gets written.

Version 2.2.0

• Support for Icecast stream metadata updates in scanning mode. When enabled,
  every time the scanner stops on a channel, current frequency is written into
  Icecast song title, which in turn is displayed in the player. Alternatively,
  textual labels can be configured for each frequency. It is possible
  to configure the amount of delay between the stream and metadata updates to
  synchronize them with the audio. There are some caveats however - read
  comments in rtl_airband.conf.example for details.
• Added global option localtime. When enabled, rtl_airband uses local time
  instead of UTC time for output file names. (Credit: ScanOC).
• Auto gain feature removed. RTL auto gain does not work well for narrowband
  channels. Most often it sets the gain too high which causes problems for
  auto squelch and audio bleeding between adjacent channels. Gain must be
  configured manually from now on.
• Dropped unmaintained Windows build.
• Reverted to power level calculation algorithm from version 2.0.2. The new
  algo didn't really do much to sensitivity, but introduced annoying clicks
  on squelch open/close.
• Improved DC offset estimator for AM mode. This one hardly ever clicks
  on squelch opening.
• Boosted AM audio volume.
• Reduced squelch flapping in NFM mode.

Version 2.1.0

• Narrowband FM demodulation support
• Automatic Frequency Control
• Append mode for recording (enabled by default)
• Dongles, channels and outputs can be individually enabled and disabled
  by a simple config flag (no need to comment out or delete large
  configuration sections)
• Use VBR for MP3 encoding
• Modified power level calculation algorithm (better sensitivity)
• Support for manual squelch setting
• Bug fixes

Version 2.0.2

• Fixed a problem with running three dongles or more, simultaneously

Version 2.0.1

• Fixed crash on output initialization

Version 2.0.0

• New libconfig-style config file format
• util/convert_cfg: can be used to convert old-style config.txt to the new format
• Syslog logging (enabled by default)
• Daemon mode
• Reworked makefiles, added install rule
• /dev/vcio is now used to access GPU on RPi; creating char_dev no longer necessary
• Startup scripts for Debian and Gentoo
• Support for auto gain setting
• Support for multiple outputs per channel
• Support for recording streams to local MP3 files
• Support for ARMv7-based platforms other than RPi (eg. Cubieboard)
• Updated documentation
• Numerous bugfixes and stability improvements

Version 1.0.0

News:

• Linux x86/x86_64 support (Windows build is currently unmaintained and might not work)
• Raspberry Pi V2 support
• Bundled hello_fft code (v2.0)
• More robust interaction with Icecast servers
• Important stability fixes