Add support for Gemini 3.5 Transcribe with SMART mode

[toorop]

Google has just released Gemini 3.5 Transcribe, including a SMART transcription mode.

Unlike standard speech-to-text, SMART mode can clean up spoken language automatically by removing filler words, repetitions and false starts, while also handling self-corrections, punctuation and formatting.

This could be particularly useful for dictation in OpenWhispr, where the goal is often to produce clean written text rather than a verbatim transcript.

Describe the solution you'd like

It would be great to add Google Gemini 3.5 Transcribe as a speech-to-text provider/model in OpenWhispr.

Ideally, users could choose between:

• VERBATIM mode for literal transcription
• SMART mode for cleaned-up dictation

Support for Gemini's custom vocabulary/context feature would also be very useful for technical terms, product names and proper nouns.

Why this could be useful

For everyday dictation, SMART mode could reduce the need for a separate LLM cleanup step.

For example, spoken input such as:

"Let's meet Tuesday... uh, no, Wednesday at two."

could directly become:

"Let's meet Wednesday at 2:00 PM."

This could improve both latency and the overall dictation experience.

Additional context

Google documentation:

Gemini 3.5 Transcribe:
https://ai.google.dev/gemini-api/docs/models/gemini-3.5-transcribe

Transcription documentation:
https://ai.google.dev/gemini-api/docs/transcribe

Google announcement:
https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5-transcribe/

I would be happy to test the integration on Windows if useful.

[toorop]

done thx