VideoLingo prompt: core/prompts.py::get_correct_text_prompt
Source revision: 01f843492759be9bbfd1a5cb697b799cc9de1609
License: Apache-2.0; see ../assets/VideoLingo-LICENSE
Modified for VidBee: named input slots and the adaptation notes below.

VidBee binding and validation:
Optional TTS pronunciation repair only; do not apply routinely to display subtitles. Keep a separate spoken-text variant. Preserve numbers, abbreviations, diacritics, names, word-internal hyphens and meaning. Skip cleaning when reducing punctuation would change meaning or pronunciation. Send a validated variant through create_dubbed_video.speechOverrides using the saved cue's one-based cueIndex and exact originalText; do not save a pronunciation variant as display subtitles. Successful tool completion, not the text proposal, proves it was used for dubbing.
The JSON-only instruction applies to this stage result, not the final user reply or tool-call envelope. Do not execute Python or make an additional provider request; apply this template within the current Agent session.

--- PROMPT ---

## Role
You are a text cleaning expert for TTS (Text-to-Speech) systems.

## Task
Clean the given text by:
1. Keep only basic punctuation (.,?!)
2. Preserve the original meaning

## INPUT
{subtitle_text}

## Output in only JSON format and no other text
```json
{
    "text": "cleaned text here"
}
```

Note: Start you answer with ```json and end with ```, do not add any other text.
