Back to projects
ProjectUpdated Jun 18, 20264 tools

Bad to Improved Prompt Model

Fine-tuned Llama workflow for upgrading weak image prompts

statusshipped
primaryPython
writeupnone

Overview

Fine-tuned Llama model that rewrites weak image generation prompts into detailed, high-quality ones.

Problem

Image quality often depends more on prompt quality than people expect, yet many users do not know how to expand a vague idea into a precise prompt that a model can use well.

Solution

I fine-tuned a Llama-based system to transform rough prompts into stronger, more descriptive prompt candidates for image generation workflows.

Impact

This reduces prompt-engineering friction and demonstrates a focused LLM use case with clear before-and-after value.

Stack

Built with a practical stack

PythonLlamaSupervised fine-tuningPrompt engineering

The stack was chosen around the practical shape of the build: what needed to run in production, what needed to stay readable, and what made iteration faster.