Most developers and technical creators know how to prompt an LLM, but few take the extra step of actually feeding it enough signal about their own voice and communication style. The gap between "AI output" and "sounds like me" is almost entirely a context and few-shot learning problem, and it's solvable without any fine-tuning, API access, or paid tools. This covers the free, prompt-layer approach to voice training in ChatGPT and Claude, including what kinds of source material work best and how to structure the context window for consistent results across sessions.
Originally published at vibrationofawesome.com
Top comments (0)