Davis-Zhang-Onehouse opened a new pull request, #20725: URL: https://github.com/apache/hudi/pull/20725
### Describe the issue this Pull Request addresses issue: #20724 Part of #20683, RFC-118 (#20684). Stacked on #20723: review only the last commit. Until the PRs below it merge, this PR also shows their commits. The audio side of speech preparation: an anchored resampler, an energy-based voice activity segmenter, and the FFmpeg audio segment source that decodes a clip's audio track into speech segments for the Whisper transcriber. FFmpeg stays optional and user-supplied, as for video. ### Summary and Changelog - `AnchoredResampler`, `EnergyVadSegmenter` and the FFmpeg audio segment source (with handle release), which turn a clip's audio track into speech segments for the Whisper transcriber. - The audio transcript flow from segments to transcript chunks. - 7 test classes. The speech test clip is regenerated by `video/generate-av-fixtures.sh` from a public-domain LibriVox recording; `LICENSE` notes it. FFmpeg stays optional and user-supplied. No code was copied from other projects. ### Impact Speech can be prepared from audio when the user supplies FFmpeg. No change outside the new source. ### Risk Level low: new code path of the new source. Verified on JDK 11 / Spark 3.5 / Scala 2.12 with 76 tests (3 skipped: they need an external Whisper model), module compilation, checkstyle and RAT. ### Documentation Update New audio settings carry their config documentation; a website page follows once the chain lands. ### Contributor's checklist - [x] Read through [contributor's guide](https://hudi.apache.org/contribute/how-to-contribute) - [x] Enough context is provided in the sections above - [x] Adequate tests were added if applicable -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
