Case study

Case Study: Migrating a Campaign Off a Deprecated Model Without Reshooting

ยท RenderBob team

A composite case study: how a studio moved a live campaign off a retiring video model without re-creating every shot from scratch.

A live campaign sequence crosses from a failing model engine to a new one while preserving character, composition, and motion.

Illustrative composite reflecting the 2026 Sora retirement pattern, not a named client.

A studio had a running campaign built on a cloud video model. Then the vendor announced the model's retirement, with the API scheduled to shut off on a fixed date. The campaign had months left to run and dozens of shots that needed to stay visually consistent with what the client had already approved. Reshooting, regenerating everything from scratch on a new model and getting fresh client sign-off, was weeks of work and a budget nobody had.

The reason it did not become a crisis was a decision made much earlier: the studio had kept its prompts model-agnostic. Each shot was described structurally, subject, motion, camera, lighting, a timed beat sheet, an audio timeline, rather than tuned to the retiring model's private syntax. The creative intent lived in portable prompts stored and versioned as assets; only a thin layer of per-model settings was model-specific.

Migration became a controlled process instead of a rebuild. The team scored the replacement candidates on the axes that mattered for this campaign, face consistency across scenes, native audio, cost per finished second, and picked the model that best matched the original's strengths. The portable prompts moved across largely intact; the work was re-tuning the thin per-model settings layer and running a shot-by-shot comparison against the approved originals to confirm the look held. Because everything was version-controlled, they could migrate incrementally, validate each shot, and roll back anything that did not match, rather than flipping the whole campaign at once.

Two supporting habits mattered. The studio ran its models behind an abstraction layer, so pointing the pipeline at a new provider was a configuration change, not a re-plumbing of the tools. And it kept cost-per-approved-shot visible, so it could confirm the new model was not quietly more expensive to reach the same quality.

The campaign continued on the new model with no reshoot and no visible break for the client. The migration cost days, not weeks. A model will get deprecated on someone else's schedule. The studios that survive it are the ones whose creative intent lives in portable assets, whose tools sit behind an abstraction layer, and whose pipeline treats every model as a replaceable component.

More from the blog

All posts