Project: Style Transfer & Brand Voice

The goal

Train a model to write marketing copy in your brand's specific voice — matching tone, vocabulary, and structure. The classic distillation use case for content teams.

Data collection

Collect 200–500 examples of approved brand copy (blog posts, social, emails).
Pair each with a "brief" (the request that produced it).
Include both good examples AND rewrites of off-brand copy → on-brand copy. The contrasts teach the model what to avoid as much as what to produce.

Evaluation is hard

Style transfer is one of the harder evaluation problems. Three approaches:

A/B blind test: have your marketing team rate base vs fine-tuned outputs without knowing which is which.
Brand voice rubric: score on specific dimensions (tone accuracy, vocabulary match, structure compliance).
LLM-as-judge with the brand guidelines as system prompt: have GPT-4o rate adherence to written guidelines.

Code

Brand-voice training example·json

{"messages": [
  {"role": "system", "content": "You are a copywriter for TechNova. Voice: confident but approachable, uses active voice, avoids jargon, prefers short sentences. Tone: optimistic, forward-looking. Never use: 'revolutionary', 'game-changing', 'synergy'."},
  {"role": "user", "content": "Write a product announcement for our new AI analytics dashboard."},
  {"role": "assistant", "content": "Meet TechNova Analytics — your data, finally making sense.\n\nWe built a dashboard that shows you what matters. No noise. No confusing charts. Just clear answers to your business questions.\n\nAsk a question in plain English. Get an answer in seconds. It's that simple."}
]}

Exercise

If your team has a brand voice document, sketch a fine-tuning brief: how many examples you'd collect, where they'd come from, what evaluation rubric you'd use. If you don't have one, take a public brand (your favorite) and reverse-engineer 5 brand-voice rules from their public copy. Same brief, just inferred guidelines.