Version 3.4.4

Maintenance release:

• Python versions less than 3.4 don't support 24-bit audio properly. We now have pure-Python shims that will allow 24-bit audio to work on those old Python versions, though they will be somewhat slower. Thanks to @danse for reporting the issue!
• Added updated Pocketsphinx binaries and Pocketsphinx installation procedures to match improvements on their end.
• Fix Unicode file paths on Windows.
• Fix caching in recognizer_instance.recognize_bing.
• We now use the Manylinux Docker image for building FLAC. Hopefully, this will make building universal Linux binaries easier for packagers.

Version 3.4.3

Bugfix release:

• Thanks to @jhoelzl, api.ai language support works again for non-English languages.

We're now GPG signing all our release tags. Under the releases page, you should see the following:

This tells you that GitHub thinks the Git tag is the same as the one we intended to release.

This key can also be found on the SKS keyservers, and you can import it with the following command:

gpg --keyserver x-hkp://pool.sks-keyservers.net --recv-keys 0x5F56B350

The packages on PyPI are signed as well - the signature can be downloaded under the "pgp" link on the SpeechRecognition PyPI page.

Version 3.4.2

Quick bugfix release on the tails of yesterday's big one:

• Add support for the monotonic library on Python 2 - if you have monotonic installed in Python 2, recognize_bing will work faster!
  • On Python 3, recognize_bing already does the things that would make it fast, so the library is unnecessary.
• Fix loading of non-16-bit AIFF files on Python 2.
• Better document the Pocketsphinx language pack installation.

Version 3.4.1

Changes:

• BREAKING CHANGE: AT&T STT API IS BEING SHUT DOWN SOON. (source)
  • For now, the recognize_att function will keep working, until the API itself is shut down.
  • It is best to transition over to IBM, Wit.ai, Google, CMU Sphinx, Bing Voice, or api.ai as soon as possible.
  • In most cases, you can simply rename recognize_att to a different service like recognize_ibm, then generate new API keys/tokens for it.
• DEPRECATED CLASS: WavFile has been renamed to AudioFile.
  • WavFile will continue to work for the foreseeable future. New code should use AudioFile.
  • AudioFile is the same as WavFile, but in addition to WAV, it also supports AIFF and FLAC files!
• New api.ai support, courtesy of @sbraden! See recognize_api in the library reference.
• New Microsoft Bing Voice Recognition API support! See recognize_bing in the library reference.
• Support for 8-bit unsigned WAV audio (thanks to @zhaoqf123 for reporting!).
• Faster, upgraded FLAC binaries, with Linux binaries using Holy Build Box for maximum distro compatibility..
• Updated setup process for Wit.ai.
• Update phrase retrieval for recognize_ibm, courtesy of Bhavik Shah from IBM.
• Documentation improvements and code cleanup.
• Clearer licensing information - see the README.

As always, you can upgrade with pip install --upgrade speechrecognition.

Version 3.3.3

Changes:

• Tiny fix to some error checking.
• The version number is all 3's now. This is pretty OK.

Version 3.3.2

Bugfix release!

• Fix exception_on_overflow shenanigans. This version will eliminate those pesky ValueErrors.
• The overflow error should well and truly be gone now.

Special thanks to @michaelpri10 for reporting the exception_on_overflow bug.

Version 3.3.1

Fix for list_microphone_names, courtesy of @ibutra. Fully compatible with 3.3.0.

See #85 for more details!

Version 3.3.0

Major changes since 3.2.1:

• Possible backward incompatibility: if PyAudio is not installed, Microphone now throws an AttributeError when created rather than not being defined.
  • This only requires changes if you are explicitly testing for the existance of the class, using something like hasattr or getattr.
• More robust error handling - always clean up PyAudio resources, no matter what error conditions arise.
• Better error checking - always verify PyAudio version.

Version 3.2.1

Changes since 3.2.0:

• Significantly improved and reorganized documentation.
• Fix for overflow issue, courtesy of @jbwincek.

Version 3.2.0

Major changes since 3.1.3:

• Support for recognition using CMU Sphinx - do speech recognition while offline!
  • English supported out of the box; French and Mandarin available for download. See the README for details.
• Automatic sample rate/sample width conversions; users shouldn't have to worry about audio formats at all.
• Lots of documentation improvements.