Dallas Crilley
Work
Live operations Shipped

Live TV production automation

Audio-aware direction for live production.

Audio-aware vMix production automation and studio tooling that runs alongside a human operator.

system
vMix auto-producer + studio toolchain
signals
mic levels, meters, show state, operator context
controls
follow speaker, avoid ping-pong, mute under rolls, return wide
ownership
Sole engineer — design, build, deployment, and operation
maintained
TV-signage control systems in production since 2019
repo
Public production-system source

The problem

Live TV production depends on fast, repeatable decisions while a human operator manages the room. Audio cues, show definitions, studio status, and signage control used to live across separate operational surfaces.

Stakeholder signal

Studio operators and production teams who need a reliable assistive layer for shot direction, show setup, and live status without surrendering the final decision to automation.

How it's built

  1. Assistive shot direction

    The auto-producer watches mic levels and vMix meters, follows the active speaker, avoids ping-pong cuts, mutes under video rolls, and returns to the wide shot when the scene needs a stable reset.

  2. Deployable show definitions

    A preset compiler turns JSON show definitions into deployable vMix bundles so the operator starts from an explicit, reviewable studio configuration instead of hand-editing a live project.

  3. Visible studio state

    The supporting studio map exposes live versus setup status across four broadcast studios, giving operators a shared read on what is ready before a production begins.

By the numbers

4 broadcast studios represented in the studio map
2019 TV-signage systems maintained in production since
human-in-the-loop operator remains in control of live decisions

Where this honestly stands today

Shipped and maintained as live production tooling. The system is designed to assist the operator with explicit shot-direction rules, recovery behavior, and a visible studio state rather than replace human judgment.

Want the parts that aren't in a public repo? Happy to walk you through the architecture and the decisions on a call.