Companion code repo for Building Speech AI (book) for Whisper learners

[prdeepakbabu]

Hi Whisper community,

I just released a book, Building Speech AI: Speech Representation, Understanding & Synthesis. A Practitioners Guide (May 2026), and the companion code repo has material that maps directly to how people use Whisper in production.

Landing page: https://prdeepakbabu.github.io/building-speech-ai
Code repo: https://github.com/prdeepakbabu/building-speech-ai

What is in the repo that is relevant here:

• examples/02_whisper_asr.py: minimal Whisper transcription with transformers and faster-whisper, including word-level timestamps via whisper-timestamped.
• examples/03_long_form.py: VAD-gated chunking (Silero VAD + 30s windows with 5s overlap) to avoid hallucinated repeats on long audio.
• examples/04_streaming.py: a low-latency streaming wrapper using whisper_streaming (LocalAgreement-2).
• speech_ai/asr.py: a thin abstraction over openai-whisper, faster-whisper, and transformers Whisper so you can A/B test WER and RTF on the same audio.

Happy to take issues / PRs on the companion repo. Thanks for building Whisper.