One-line positioning

Produced AI character plates for The Wizard of Oz at Sphere, with all assigned outputs delivered into the final immersive film pipeline.

Context

This project rebuilt legacy Wizard of Oz character moments for Sphere-scale playback with a large cross-functional team. Shot responsibilities were distributed by sequence, and handoffs moved across preprocessing, generation, and composition functions.

My role

I handled assigned character coverage and generation execution: Good Witch and Bad Witch medium shots, Bad Witch flying monkeys, large Munchkin sets, Bad Witch guard shots, and one Scarecrow close-up. My tasks included per-character cutout prep, image upscaling with Terra plus LoRA for degraded subjects, motion-matched video generation, and color-matching checks before comp.

Team setup

  • Large cross-functional team with shot-based task distribution.

  • Pipeline included source processing, AI generation, and composition functions.

  • Ownership model: generation leads stayed on assigned shots until they were usable in final edit.

Workflow

  • Original footage processing: segment characters from source and isolate them from background for downstream generation.

  • Image upscaling and baseline correction: recover degraded subjects (especially Munchkin material) with Comfy UI and LoRA, then run base color correction.

  • Video generation: generate shots from segmented and upscaled inputs while preserving original action timing and choreography using google VEO.

  • Post-composition handoff: deliver generated material to composition for final integration.

  • Image detail optimization: perform large-volume still upscales, compare against source frames, and finalize lighting/color parity checks.

Key deliverables

  • Good Witch and Bad Witch medium-shot assets.

  • Bad Witch flying monkey asset sets.

  • High-volume Munchkin character assets and dance-related shot material.

  • Bad Witch guard assets and one Scarecrow close-up sequence.

Technical challenge

The hardest constraints were action accuracy and rhythm sync. Generated motion had to match original film choreography exactly, and Munchkin dance groups had to stay fully on beat with music timing. At the same time, upscaled characters needed lighting and color parity with source frames to avoid mismatch at Sphere scale.

Outcome

Delivered production-ready assets across all assigned shot categories, and all delivered shot outputs were used in the final film.