Whisper supports an initial_prompt parameter that biases the model's vocabulary during transcription. For Whisper models, we now join the user's custom words list into a comma-separated string and pass it as the initial_prompt. This guides the model to prefer those spellings during decoding rather than relying solely on fuzzy post-correction.

For non-Whisper engines (Parakeet, Moonshine, SenseVoice, GigaAM), the existing apply_custom_words post-correction continues to apply since those engines don't support prompt-based vocabulary biasing.

Before Submitting This PR

Please confirm you have done the following:

• I have searched existing issues and pull requests (including closed ones) to ensure this isn't a duplicate
• I have read CONTRIBUTING.md

If this is a feature or change that was previously closed/rejected:

• I have explained in the description below why this should be reconsidered
• I have gathered community feedback (link to discussion below)

Human Written Description

From the discussion:

I've encountered an issue with custom vocabulary/custom words in Handy. The current bigram replacement mechanism can be overly aggressive - it sometimes replaces words that shouldn't be replaced. However, removing those entries from the custom words list means the desired words never appear at all.

I know that Whisper supports an "initial prompt" concept, which biases the model toward producing specific words without forcefully replacing output. While looking through the codebase, I noticed there appears to be a way to use custom words to generate an initial prompt, and I wanted to ask whether a pull request for this would be welcome.

Related Issues/Discussions

Fixes #
Discussion: #1009

Community Feedback

Testing

No unit tests for this. I built and used this on my machine. The use case I was working on was "CLI" and "call" - that was fun to see in transcripts once I noticed it :)

Screenshots/Videos (if applicable)

AI Assistance

• No AI was used in this PR
• AI was used (please describe below)

If AI was used:

• Tools used: claude-code opus 4.6
• How extensively: I knew of the capability from prior research and came up with the idea of using the existing vocabulary as the prompt instead of replacement. I then used claude-code to see if the fix was small and implement it. I then reviewed it, but I admit I have not written rust before, so my human review may be insufficient.