Nex Study MusicAn AI Music and Video Channel
A YouTube channel of hour-long focus mixes where the music, the looping visuals, the titles and the thumbnail all come out of one AI pipeline, with every track mastered to the loudness top study channels use.
Built with Lyria 3 Pro, Gemini image models, LTX-2.3, Kling 3.0, Python, FFmpeg, torchvision pose model, YouTube Data API.
1 hour
Length of the first public mix, music and visuals both generated
-14 LUFS
Every mix mastered to the loudness of the top study channels
3 skills
Music, visuals, and packaging, each run by an AI agent
What it is
Nex Study Music is my own channel for deep-focus music, and a working test of an AI production pipeline. One skill makes the music: Google's Lyria 3 Pro is the default engine, with a local open model and a code synthesizer as fallbacks, and every mix is mastered to -14 LUFS. A second skill makes the visuals: scene stills from Gemini image models, animated in code into two-minute loops, plus character shots from video models anchored on both a start and an end frame so they do not drift. A third skill writes the title, description and thumbnail and assembles the upload manifest. The first public mix, 'Ocean Dream', is a one-hour underwater ambient set published on 2026-10-28.
The problem
Study-music channels need hours of new audio and visuals every week. Generated music often sounds thin next to the channels people already play, and generated video loops jump at the seam or let a character's face and body change between shots.
What I built
The music was measured against the top five study channels and rebuilt to match: a sub layer, lower chords, a slow swell and a -14 LUFS master. Visuals are built from still scenes animated in code, with short generated clips crossfaded into seamless loops. Clips are anchored on start and end frames, and running scenes are checked with a pose model for a believable stride. Every paid generation run starts with a cost estimate.
What happened
The channel is live with its first public one-hour mix, published on 2026-10-28.
What it changed
- Shows the AI music and video work behind the NexAI production service on a public channel
Who it fits
- Brands that want original background music and visuals for videos and stores
- Creators who want a steady stream of long-form ambient content
- Music engine order with fallbacks: Lyria 3 Pro, a local open model, then a code synthesizer
- Seamless video loops crossfaded from short generated clips
- Start and end frame anchors that stop drift in generated character shots
- Pose-model check of stride rate on generated running scenes
- Titles, descriptions and thumbnails written by the same pipeline
Want something like this built?
Bring the job that eats your week. Step 1 is a free 15-minute fit call. Step 2 is the AI Staff Assessment, $500, a written assessment you keep. Step 3 is your first hire at a published price, or you take the assessment and stop.