AI Polish Prompt Builder
Build a reusable prompt for voice transcripts, then verify it with the same sample text
Polishing Style
Custom Dictionary
One term per line. These will always be spelled exactly as written.
Extra Instructions
Any additional rules for the AI.
Generated Prompt
You are a professional text editor. The user will provide raw voice transcription text. Your task: 1. Fix all grammar and punctuation errors 2. Remove filler words (um, uh, like, you know, basically, etc.) 3. Improve sentence structure for clarity 4. Maintain a professional, formal tone 5. Do NOT change the meaning or add new content 6. Output only the polished text, no explanations Safety and fidelity rules: - Preserve names, numbers, dates, URLs, code, quotations, and custom dictionary spellings - Do not invent facts, commitments, citations, or missing details - When the transcript is ambiguous, keep the uncertain wording instead of guessing - Output only the transformed text unless the user explicitly asks for another format
Paste this into the System Prompt field in OpenTypeless Settings → AI Polish → Custom Prompt.
FAQ
Where do I paste this prompt in OpenTypeless?
Go to Settings → AI Polish → Custom Prompt. Paste the generated prompt into the system prompt field and save. OpenTypeless will use it for all subsequent voice input polish operations.
Can I use different prompts for different tasks?
Yes. Generate a version for each workflow, save dated copies externally, and paste the version you want into the Custom Prompt field.
How should I compare providers or models?
Use the same raw transcript, prompt, language, and review checklist. Compare fact preservation, formatting, latency, and failure behavior, then keep the result that fits your workflow.
What this prompt builder produces
The browser produces a reusable instruction prompt. Copy it to your selected service or desktop polishing settings to run the actual text transformation. Review that service’s costs and data terms before testing a transcript.
- Specify the desired tone and output format.
- Preserve identifiers, figures, names, and any quoted text.
- Copy the generated instruction and test it with a short original transcript.
- Compare the result with the original before adopting it as a default.