v1.4.0

• whisper_cpp_vendor: whisper.cpp 1.6.2 to 1.7.2 release, build changes
• Added live audio transcription streaming
• whisper_server:
  • Holding incoming Audio data in a Ring Buffer (removed BatchBuffer, drop oldest audio).
  • Transcribing the entire buffer of audio data with whisper.cpp on a timer interrupt
  • Publishing the resulting tokens + probabilities on topic /whisper/tokens
  • Removing the Action Server
  • New Node Parameters:
    • active -- Boolean to control if whisper.cpp should be run or not.
    • callback_ms -- Integer controlling how often whisper.cpp is called.
    • buffer_capacity -- Integer number of seconds previous where audio is transcribed.
• transcript_manager package added:
  • Store record of what was previously transcribed.
  • Track what is currently being transcribed. Align and update the text from subscribed topic /whisper/tokens.
    • Updates done on timer interrupt
  • Host the Action Server which was previously part of whisper_server
  • Publish the entire transcript (previous and current) under /whisper/transcript_stream
    • Published transcript contains text and estimated segment markings, segment timestamps
• whisper_demos: Add stream node
• whisper_idl:
  • Added msg/WhisperTokens.msg, msg/AudioTranscript.msg
  • Added launch/replay.launch.py which does not bring up audio_listener
• whisper_util: Changes to directly inference and then serialize whisper.cpp model output, also containing probability data.

Also refer to https://github.com/ros-ai/ros2_whisper/blob/main/CHANGELOG.rst. Work done by @NathanCorral

v1.3.1

• whisper_msgs: Changed to whisper_idl package
• whisper_bringup: Changed executor to MultiThreadedExecutor so audio and inference can run in parallel on whisper_server

v1.3.0

Some update to the vendor package as well as downstream updates to parameters

• whisper_cpp_vendor: whisper.cpp 1.5.4 -> 1.6.2 release
• whisper_cpp_vendor: CMake build flag WHISPER_CUBLAS to WHISPER_CUDA
• whisper_server: Removed launch mixins
• whisper_server: Updated parameters for whisper.yaml
• whisper_server: Fixed whisper initialization order

v1.2.1

• Changes whisper_cpp_vendor C++ version to C++11

v1.2.0

• Now supports full CUDA backend with whisper.cpp v1.5.0, refer https://github.com/ggerganov/whisper.cpp/releases/tag/v1.5.0
• Fixes small bugs

v1.1.0

Refer to #6

• Improved action server for whisper_server
• Improved terminal output for whisper_demos

v1.0.0

Refer to #3

alpha-v0.1.0

• basic inference using whisper
• custom audio listener implementation