Afrikaans learner: Himalayan onions (unscripted)

In this 3.45-minute, 38-sentence audio, a native English speaker tries to learn Afrikaans through a conversation with Gemini about a plant that Gemini identifies as a Himalayan onion. It supplies the speaker with a recipe that is near impossible to understand due to the presence of words from the contaminated Afrikaans dataset that Gemini is only able to spell out. There are three words from the bad dataset used, one of them in two different forms, repeatedly.

Every package is hand-curated by a native speaker, and includes:

 - original audio.wav

 - raw Whisper X and Faster Whisper output

 - human edits - transcripts and English translation in:

      - .txt

    - .json

    - .csv

 - ML package with accurate timestamping and perfect 

  single-sentence cuts side-by-side with human-curated transcript snippets.

 - detailed, human-curated manifest

 - diff report

 - speaker profile

 - license