What it was

Audizer started in 2011 as a way to automate the slow, repetitive parts of music production — the steps a working engineer takes between tracking and a finished master. The first version was built on RPA and cloud service automation: stitching together the existing tools and DAW workflows engineers already used, then running them headlessly so a session could be processed without a human babysitting every parameter.

Years later, Audizer was rebuilt around large language models. Instead of hard‑coded automations, a master channel acted as the conductor: it analyzed the full mix, produced structured output describing what each track needed, and dispatched optimization suggestions back to per‑instrument plugin children — drums, guitar, bass, vocals, and so on. User feedback flowed back through the master channel, the LLM updated its plan, and the children re‑tuned themselves accordingly. The result was a mix engineer that could reason at the song level instead of the knob level.

What it did

• Headless mixing and mastering automation via RPA against existing audio tools (2011 era)
• Cloud rendering pipelines for batch processing of stems and full sessions
• Master‑channel orchestrator that analyzed a mix as a whole and produced structured optimization plans
• Per‑instrument plugin children (drums, guitar, bass, vocals, keys) that received and applied targeted adjustments
• Feedback loop: user notes — "more punch on the kick", "vocal is too loud in the chorus" — routed back through the master channel and into the relevant children
• Structured output between the master and the children so coordination was deterministic and inspectable
• Native VST plugin built in JUCE (C++) so the LLM‑driven version ran inside any DAW the engineer already worked in

Technologies

Status

Audizer is not currently operational. The architecture, plugin scaffolding, and processing pipelines are preserved.