Finding where AI actually fits in a VFX and post-production pipeline
In November 2022, as ChatGPT was introduced, a post-production studio working across VFX, editing, and motion graphics — for film, TV, streaming, and advertising, as well as pre-production work like pre-visualization, concepts, pilots, and pitch decks — reached out after conducting their own preliminary research on it. Studios that couldn't offer AI-assisted services could soon lose work to ones that could, , and as far as we knew, no studio in the Indian market had worked out how to do this responsibly. Our Risk was: Applying AI tools without understanding where they fit could damage the quality of the work and the credibility of the studio.
Over several months, I was working alongside them on multiple client-facing projects while continuously testing tools in small, low-stakes ways as opportunities arose. This research was paired with a review of the broader tool landscape and how (or whether) other studios were approaching AI integration, resulting in a set of practical guidelines for where AI could responsibly contribute to the studio's workflow.
They had reached out proactively about AI's potential. The question wasn't whether to adopt, but where it could help without undermining the quality of work.
Early, undirected use of AI tools showed clear limits: full-scope generation broke down on style consistency and produced assets that needed significant rework. Most tools, including paid ones, required regenerating an entire asset just to fix one detail.
Meanwhile, clients frequently demanded large volumes under tight deadlines, straining teams. This suggested that AI's real value might lie in absorbing specific, well-scoped tasks.
Rather than testing in isolation, I worked inside the studio's actual production environment, observing across real, live client work.
This surfaced four guiding principles: directing AI at specific, well-defined tasks produced far more usable results; AI was best suited to absorbing high-volume, low-stakes work — like frequently revised background variations; applying AI indiscriminately across the entire pipeline wasted render time, so its use had to be scoped to specific phases; and tools capable of isolated, localized edits were far more usable.
That last principle pointed at Adobe Generative Fill as the framework's best option — not only because it supported isolated edits, but because the studio already held a studio-wide Adobe Creative Cloud subscription, making it suited for long-term adoption, while more specialized tools could be used on a case-by-case basis.
The studio was approached to produce a movie pre-visualization/concept video for a director — an opportunity to test under real, live-client conditions.
Read Case StudyEmbedded research and use-case discovery, alongside live production work.
Four core guidelines were established for responsible AI integration.
The first framework of its kind identified in the Indian post-production market at the time.