Open to interesting work
I’m Drew, an open-source developer working where AI, audio, and useful tools meet. My projects turn difficult technical workflows into products anyone can run.
Selected work / 01 to 06
Products spanning generative audio, local AI, media infrastructure, developer tools, and delightfully specific problems.
Turn ebooks into fully chaptered audiobooks with voice cloning and support for 1,704 languages, all running locally on CPU or GPU.
Hear the output
Real samples generated with ebook2audiobook, from the default narrator to cloned, fine-tuned, and multi-character voices. Press play to hear the finished result.
A longer-form sample of the voice included out of the box.
Compare a standard single-narrator conversion with automatic narrator and character voice assignment from E2A-SML.
Generated voice demonstrations are presented for technical comparison. They are not affiliated with or endorsed by any person whose vocal characteristics inspired a sample.
Inside the E2A ecosystem
I built a synthetic-data-assisted training pipeline around the Universal TTS Finetuner: clone a voice from a few seconds of reference audio, generate hours of clean teacher data with XTTS, then train a compact Piper model that runs faster and on far less hardware.
Preserving an unusually quiet vocal style through a compact model.
A recognizable narration style transferred into a lightweight runtime.
A unified training environment for building and adapting speech models, including the teacher-to-Piper workflow demonstrated above.
Explore the finetuner ↗Analyzes books, attributes dialogue to characters, and produces voice-tagged SML for multi-speaker audiobook generation.
Explore E2A-SML ↗Research demonstration. Generated voices are presented for technical comparison and are not affiliated with or endorsed by the people whose vocal characteristics inspired them.
A desktop app and CLI that turns Jellyfin libraries into polished DVD sets: menus, subtitles, disc splitting, artwork and all.
View project ↗A desktop noise-evidence system that detects loud and low-frequency events, captures audio and video, and logs calibrated measurements for reporting to local authorities.
View project ↗
THE IDEA LIVES ON INThe original multi-character audiobook project: it detected characters, attributed dialogue, and assigned distinct voices automatically. Its successor now lives inside ebook2audiobook as E2A-SML.
01:42:18
“…the exact moment you were looking for.”
03:17
Semantic match across your library
Download, transcribe, and semantically search entire video libraries locally using Whisper, FAISS, and embeddings.
View project ↗A Go-powered terminal music player that browses, streams, and syncs a remote Apple Music library over SSH.
View project ↗Turns a Jellyfin library into always-on virtual TV channels with intelligent scheduling, filters, and era-appropriate ads.
View project ↗About / The short version
My best work starts with “why is this still so hard?” and ends with a tool people can actually install.
I work in IT at Adult Swim while studying at Georgia State University, where I am also a member of the Alser Lab in the Department of Computer Science.
I build in public, listen to users, and follow projects past the prototype stage into packaging, documentation, releases, and long-term maintenance.
Information technology
Student in Atlanta, Georgia
Computer science research
Visit the lab ↗
Recognition / Ebook2audiobook
Ebook2audiobook earned 1st Place in Applied Research and Entrepreneurship for its poster presentation and a Global Engagement Award at the Georgia State Undergraduate Research Conference.
Project sponsor: Professor Mohammed Alser
View the official winners ↗Have an interesting problem?