v0.5.1

What's Changed

• Backslash missing in example by @t-nil in #236
• Making backend arg safer by @berkaybilik in #243
• Pin silero VAD onnx model version to v4.0 by @makaveli10 in #247
• Fix deadlock issue in FFmpeg subprocess by ensuring stderr is consumed by @sondt2709 in #257
• Discard no speech segments by @makaveli10 in #262
• Initialize last_segment to None by @makaveli10 in #272
• Set compute_type based on device capability by @makaveli10 in #274

Full Changelog: v0.5.0...v0.5.1

v0.5.0

What's Changed

• Improve cpu and gpu Dockerfiles, resulting in much smaller images by @peldszus in #206
• Add option: save network stream to local file while transcribing by @fraic in #192
• fix: limit CPU usage for VAD onnxruntime inference session by setting… by @makaveli10 in #215
• Add support for RTSP stream by @dshepelev15 in #212
• Make writing audio frames optional by @makaveli10 in #216
• Expose the srt file location of Transcription client by @chien-liu in #224
• Update tensorrt llm to v0.9.0 by @makaveli10 in #227
• Single model mode by @peldszus in #223

Full Changelog: v0.4.1...v0.5.0

v0.4.1

What's Changed

• Don't clear server_error flag on close. by @jsichi in #187
• Fix client close by @makaveli10 in #190

Full Changelog: v0.4.0...v0.4.1

v0.4.0

What's Changed

• build & push docker image on every new tag by @makaveli10 in #161
• Upgrade faster whisper==1.0.1 by @makaveli10 in #163
• Handle failure on systems without microphones by @FlippFuzz in #173
• Fix faster whisper version in setup.py by @FlippFuzz in #175
• Add support for processing same audio stream via multiple clients running different tasks. by @jsichi in #174

New Contributors

• @FlippFuzz made their first contribution in #173
• @jsichi made their first contribution in #174

Full Changelog: v0.3.0...v0.4.0

v0.3.0

What's Changed

• Fix docker image gpu by @makaveli10 in #156
• fix: lanuguage, task prefix in decoder start ids by @makaveli10 in #157
• fix: cpu usage issue by @makaveli10 in #158

Full Changelog: v0.2.1...v0.3.0

v0.2.1

• Add unittests
• update requirements to include scipy
• code format/quality check according to flake8
• vad is now an option that the client can set if they want to use it or not
• docker image for both cpu and gpu are pushed to ghcr on new tag

v0.2.0

• Add unittests
• update requirements to include scipy
• code format/quality check according to flake8
• vad is now an option that the client can set if they want to use it or not
• docker image for both cpu and gpu are pushed to ghcr on new tag

v0.1.0

Major

• TensorRT Backend
• Save Trasncript as srt file at the end of session

Minor

• Fix key error for initial_prompt and vad_params
• Change model size param name
• Custom model model usage controlled by server

v0.0.11

• bug fixes

v0.0.10

• Support large-v3
• bug fixes