v0.7.10

What's Changed

• Update FluidAudio version to 0.7.9 by @BrandonWeng in #192
• Add word-level timestamps support to CLI transcribe command by @Alex-Wengg in #193
• optionalize TTS via FluidAudioTTS target by @Alex-Wengg in #186
• Expose streaming chunk API and package ESpeakNG xcframework with dSYMs by @Alex-Wengg in #201
• Streaming Diarization Improvements by @SGD2718 in #191

Full Changelog: v0.7.9...v0.7.10

What's Changed

• Update FluidAudio version to 0.7.9 by @BrandonWeng in #192
• Add word-level timestamps support to CLI transcribe command by @Alex-Wengg in #193
• optionalize TTS via FluidAudioTTS target by @Alex-Wengg in #186
• Expose streaming chunk API and package ESpeakNG xcframework with dSYMs by @Alex-Wengg in #201
• Streaming Diarization Improvements by @SGD2718 in #191
• Fix: Move ESpeakNG.xcframework to top-level Frameworks directory by @Alex-Wengg in #205

Full Changelog: v0.7.9...v0.7.10

v0.7.9

What's Changed

• MacOS catalyst support by @BrandonWeng in #181
• add @available(iOS 17.0, *) for VBxClustering and OfflineEmbeddingExt… by @Josscii in #182
• SpeakerManager improvements by @SGD2718 in #180
• Reorganize mac catalyst bundle by @BrandonWeng in #183
• Add Speakmac to showcase by @kiranjd in #187

New Contributors

• @Josscii made their first contribution in #182
• @SGD2718 made their first contribution in #180
• @kiranjd made their first contribution in #187

Full Changelog: v0.7.8...v0.7.9

v0.7.8

Impact:

• Remove shared buffers for diarization pipeline that was causing concurrency crashes, < 3% impact to latency.
• Reduced missing words by 10% when running ASR on long audio files
• Slightly improved WER for v2 and v3 (~0.5% on benchmarks) and ~5% faster!
• Programmatically override the default registry to download from. (i.e hf-mirror.com), useful for Chinese developers

What's Changed

• Make ANE Utils concurrency safe by @BrandonWeng in #172
• Standardize registry override by @BrandonWeng in #175
• Fix Outdated speakermanager doc by @Alex-Wengg in #176
• Switch ASR to stateless for batching by @BrandonWeng in #177
  • shoutout to @hamzaq2000 for his help here!

Full Changelog: v0.7.7...v0.7.8

v0.7.7

What's Changed

• Add Intel ESpeakNG support so its easier for others to build by @BrandonWeng in #167

Full Changelog: v0.7.6...v0.7.7

v0.7.6

What's Changed

• Remove unsafe flags from SPM by @BrandonWeng in #166

This should fix SPM issues

#164

Full Changelog: v0.7.5...v0.7.6

v0.7.5

What's Changed

Core ASR & Diarization

• Fixed VAD threshold overriding per segment (#153, #155) — thanks @starcrest
• Added pyannote community-1 model for offline speaker diarization (#150) — @BrandonWeng

Speech Synthesis (ESpeakNG)

• Fixed ESpeakNG framework structure (resolves #159, #160) (#161) — @antonlvovych
• Added ESpeak linking tests (#162) — @BrandonWeng

Dataset & Pipeline Improvements

• Expanded FLEURS dataset coverage to all 25 languages
• Added Hugging Face download retries for robustness (#158) — @BrandonWeng

Internal Cleanups

• General refactors and organization improvements across diarization and data pipelines.

New Contributors

• @starcrest made their first contribution in #153
• @antonlvovych made their first contribution in #161

---

Full Changelog: v0.7.4...v0.7.5

v0.7.4

What's Changed

• Fix Kokoro File phonic not found issue with xFramework by @BrandonWeng in #151

Should build and run properly now, verified with another developer

Full Changelog: v0.7.2...v0.7.4

v0.7.2

What's Changed

• Update ESpeak and hard fail if missing by @BrandonWeng in #148
• Include header for import files by @BrandonWeng in #149
• Bump min versions to macos 14 and iOS17. This shouldn't matter for most as our models are built against MacOS14 and iOS17..

Full Changelog: v0.7.1...v0.7.2

v0.7.1

What's Changed

• Adding support for phonetic and alias replacement by @smdesai in #140
• Fix ML cache race condition for streaming ASR by @BrandonWeng in #147

New Contributors

• @smdesai made their first contribution in #140

Full Changelog: v0.7.0...v0.7.1