Hi everyone! 👋 I’m still pretty new to voice cloning and just finished my first project with Resemble AI. I hit a snag that feels obvious now, but I totally missed it at first.
I recorded about 20 short training clips, but I wasn't careful about how close I was to the mic. Some clips were much louder than others. The resulting voice model sounded a bit… uneven and glitchy in places. A more experienced friend asked if I'd normalized the audio levels before uploading. I hadn't! Once I leveled the volume across all clips and retrained, the quality improved a lot. The delivery was much more consistent. Just wanted to share in case any other beginners are about to upload their dataset! Has anyone else run into this? What's your usual process for prepping audio?