I recently found myself with 252,000 ElevenLabs credits and very little experience using the platform. Naturally, I decided to put those credits to work for my publishing business, Pen Odyssey. My AI collaborator, Quill, and I began exploring the possibilities. We initially considered producing a three-character conversation featuring our narrator, Father Goose, and The Cat. Then reality intervened. Lol. The platform was unfamiliar, the navigation wasn't always intuitive, and the voice library offered more than 10,000 choices. So we changed our approach. Here are four decisions that made a difference. 1. Write the actual script before choosing a voice. We discovered that ElevenLabs suggested voices based on the text we entered. Instead of endlessly searching through thousands of possibilities, we let the material narrow the choices. We auditioned promising voices and chose one for each role. 2. Learn to walk before attempting a marathon. Our original plan involved multiple characters speaking in one production. We postponed that. Instead, we began producing useful, single-voice recordings. That allowed us to learn the basic tools while creating finished assets for our business. 3. Batch recordings by voice. Constantly switching voices seemed unnecessary. We wrote and approved several scripts for each selected voice, then generated all those recordings before moving to the next character. One voice. Multiple scripts. One production session. 4. Standardize the workflow and file organization. We established a simple process: Write → Review → Approve → Select Voice → Generate → Download → Archive. We also adopted descriptive filenames and a preferred MP3 format. One surprise: changing the download quality setting after generation didn't improve an earlier recording. Selecting the desired quality before generating subsequent recordings produced the expected results in our tests. The results? We've now created 15 reusable audio recordings across three voices: