Work & case studies

Pipelines, apps and tools I've built and run in production. Big media lives on GitHub — linked here, not bloating the page — so this stays fast.

FILM / TV · CTV Original · Commissioned by Urban Post

“Acting Good” — AI Facial Animation for a CTV Original

For the CTV Original comedy series Acting Good (CTV Comedy / APTN, streaming on Crave & APTN+), commissioned by Urban Post: from ARRI 35 LogC4 raw footage and DPX frames, I drove a character’s facial animation to speak six scripted lines using slug audio for lip sync.

  • · CTV Original series — airs on CTV Comedy, streams on Crave & APTN+
  • · ARRI 35 LogC4 raw footage processed from DPX frames
  • · Slug (temp) audio used to drive facial animation + lip sync
  • · Six scripted dialogue lines animated for a broadcast episode
  • · Delivered into a real Urban Post finishing workflow
CTV OriginalARRI 35 / LogC4DPXLip SyncFacial Animation
Watch the reel

PIPELINE · Proprietary IP · Enterprise / NDA

Air-Gapped Local AI Production Pipeline

A 100% offline automation pipeline for studios under strict NDAs — generation, voice and video all processed locally with no internet or cloud-API dependency.

  • · Text→Image, Text→Video, Image→Video, Video→Video, video extension
  • · Qwen3-TTS voice cloning + Wan Multitalk lip-sync dialogue
  • · Character sheets, local LoRA training, scene manipulation
  • · Runs on a local loop (127.0.0.1:8188) — zero telemetry, zero-trust
ComfyUIFLUXWAN 2.2LTX 2.3Qwen3-TTSRTX 5090
ComfyUI node graph for a local, air-gapped AI generation pipeline
ComfyUI graph — dance motion transferComfyUI graph — voice cloningComfyUI graph — lip-sync dialogueComfyUI graph — image-to-animationComfyUI graph — agent tooling

APP · Shipped app · Live on the App Store

Langly — Learn Brazilian Portuguese

A 100% offline language-learning app I designed, built and shipped to the App Store myself — four modules (vocabulary, common sentences, pronunciation, Q&A) with custom word blocks and no account to create. Built end-to-end with AI-assisted development, from first commit to App Review.

  • · Live on the App Store — free, iOS 18+, no sign-up required
  • · Four modules: vocabulary, common sentences, pronunciation, Q&A
  • · Fully offline — no cloud calls, nothing leaves the device
  • · Custom word blocks + audio mode for hands-free practice
  • · Designed, built and shipped end-to-end by one person
SwiftUISwiftDataStoreKitAVFoundationApp Store
View on the App Store
Three Langly app screenshots — main menu, vocabulary and pronunciation modules

APP · Shipped app · Flask / Whisper

ReMyk — Remote Voice Dictation

Browser-based remote dictation that runs entirely over your LAN. Point your phone (or any mic) at the relay and your Mac Studio transcribes it — no cloud, no install, ~2-second round trip.

  • · Browser mic → LAN relay → local Whisper AI transcription
  • · Zero-install, self-hosted, private by design
  • · Built for hands-free, eyes-up work on set and at the desk
PythonFlaskWhisper AIWeb Audio API

TOOL · Shipped tool · Colour science

DIT Ungrade Tool

Reverses on-set grades back to camera LOG so you can build honest portfolio comparison sliders — the tool I wanted on every set, so I built it.

  • · Alexa / Alexa 35 / Sony Venice · K1S1 LUT support (ported from LogC)
  • · Python + tkinter GUI, built for real DIT workflow
  • · Colour-accurate, repeatable, no creative suite required
PythonColour ScienceArri / Sonytkinter

Renders & character work

A sample of images and character sheets produced in the local, air-gapped pipeline.

Character reference sheet (Ryan) — wide formatCharacter reference sheet (Vee)Character reference sheet (Vee) — wide formatGenerated still — render 01Generated still — render 02Character reference sheet (Ryan)Character reference sheet (Ryan) — alternate

More builds live on GitHub

Tools, experiments and the portfolio itself — the repo is the source of truth.