Exploring the process of using large language models (LLMs) to generate creative lyrics, this post outlines several techniques and methodologies to improve the quality and creativity of the outputs. It begins with simple prompt requests and evolves through structured approaches such as the 'lyrical factory' and the 'loosey-goosey duo.' The journey highlights the use of 'reminding the model' technique, Plan + Execute pattern, and few-shot prompting, supplemented by human oversight and a semantic store of existing lyrics. Despite significant improvement, the quest for high-quality AI-generated lyrics continues.
Table of contents
Part 1: A simple song writerPart 2: A lyrical factoryPart 3: The loosey-goosey duoPart 4: Just add humanPart 5: A few-shot rethinkPart 6: Not the endShare this:Like this:Sort: