Skip to main content
“og:title”: “Stenox Docs”

What are System Prompts?

System prompts tell the AI enhancement provider how to process your transcribed text. They control grammar correction, formatting, tone, and structure.
Each profile can have its own system prompt — use professional tone for work emails and casual tone for personal notes.

Default Enhancement

Both Stenox Cloud and local providers come with fine-tuned enhancements that we continue to refine. No configuration needed — Stenox works great out of the box.
Custom prompts are optional for general use. However, profiles with custom prompts are powerful for users who want specific behavior per app — like professional tone in email clients or technical formatting in code editors.

Customizing Your Prompt

1

Open Profile Settings

Settings → Profiles tab → Select your profile → Click Edit
2

Enable LLM Enhancement

Toggle on Enable LLM Enhancement in the Post-Processing (LLM) section
3

Find System Prompt

Scroll to the System Prompt field — labeled “Instructions for the AI on how to format text”
4

Edit Prompt

Enter your custom instructions in the text box
5

Save Profile

Click Save Profile to apply changes

Example Prompts

Professional Email

Casual Notes

Technical Documentation

Meeting Notes

Creative Writing

Medical/Legal (Minimal)

Prompt Tips

Instead of just “improve this”, specify the tone you want:
  • “professional and formal”
  • “friendly and conversational”
  • “technical and precise”
  • “casual and relaxed”
Tell the AI what to preserve:
  • “Keep technical terms as-is”
  • “Don’t change proper nouns”
  • “Preserve my original phrasing when possible”
Request specific formatting:
  • “Format as bullet points”
  • “Add paragraph breaks”
  • “Use markdown headers”
  • “Keep as a single paragraph”
Control how much the AI changes:
  • “Make minimal changes” — Light touch
  • “Polish and improve” — Moderate editing
  • “Significantly rewrite for clarity” — Heavy editing

Prompt Patterns

Minimal Editing

Moderate Editing

Heavy Editing

Per-Profile Prompts

Different contexts need different prompts: Learn about creating profiles →

Provider Differences

Different AI providers interpret prompts differently:
For local MLX models: Keep prompts simple and direct. Complex multi-step instructions may not work as well as with larger cloud models.

Testing Your Prompts

After changing a prompt:
  1. Test with typical input — Dictate something you’d normally say
  2. Check the output — Does it match your expectations?
  3. Iterate — Refine the prompt until it works consistently
  4. Edge cases — Test with unusual input to catch issues

Troubleshooting

Add constraints:
Be more directive:
Be explicit about format:
Or:
Different models have different capabilities. Simplify prompts for local MLX models, or create separate profiles for different providers.

Next Steps

Create Profiles

Set up different prompts for different contexts.

Choose Provider

Pick the right AI provider for your needs.