Lesson 4 of 4

Iterating on Voice

8 min
expert

Training AI on your voice is iterative. You'll generate output, evaluate it, give feedback, and generate again. Each cycle should get closer to your authentic voice. Here's how to iterate efficiently.

Key Concepts

  • 1
    Generate → Evaluate → Feedback → Regenerate
  • 2
    Be specific in feedback: 'more punchy' vs 'shorter sentences, more verbs'
  • 3
    Save successful outputs as reference examples
  • 4
    Discard and restart if output is fundamentally wrong direction
  • 5
    Patience: good voice matching takes multiple sessions

Examples

Prompt / Technique
Iteration example:
V1: Too formal, stiff
Feedback: 'Make dialogue more casual, characters interrupt each other'
V2: Casual but lost tension
Feedback: 'Restore tension through shorter sentences in climactic moment'
V3: Close—adjust one character's vocab
Result / Insight

Each iteration gets closer to target voice

Pro Tips

Don't try to fix everything at once—work on one element per iteration
When something works, note exactly why
Sometimes the issue is your instructions, not AI's output
Know when to cut losses and write it yourself

Summary

Voice matching is iterative. Generate, evaluate, give specific feedback, and iterate until output feels like yours.

Swipe left/right to navigate lessons

Continue Learning