How do you use AI to extend a small studio set into a larger environment?

Extend a small set by treating your practical shot as the anchor: upload screenshots from your own shoot as visual references, then generate wide environment shots that place the same action in a larger world. Keep AI on wides, splice the best seconds from multiple generations, and blend with color grade and sound design. One production extended a 6x6 ft studio pool into an open ocean this way.

Start by shooting your practical anchor shot — the set still needs to be built and lit, because AI extends physical production rather than replacing it. In one documented production, the team constructed an 8x8 ft cave wall and a 6x6 ft studio pool, then used AI to build the world around both.

1. Upload screenshots from your practical footage as references. Pull stills from your shoot and load them into the invideo agent before generating anything — invideo is an agentic video creation tool with the current generation models available, and it uses your stills to anchor color, contrast, and lighting to your real material. Reference-based generation is the single biggest realism lever: text-only prompts produce environments that won't match your set's look.

2. Generate wide environment shots outward from the locked shot. Ask the invideo agent for wides that place the same action in a larger world — the 6x6 ft pool was too small to hide set edges, and the actor could tread water for only one to two minutes per take, so the production generated wide red-ocean and shoreline shots around the tight practical footage. Where model choice matters, the invideo agent routes the job: Seedance 2.0 reference-to-video carries your still's look into moving environment shots, while Veo and Kling handle expansive establishing wides — all available inside invideo, so you don't pick a platform per model.

3. Keep AI on wides; keep your camera on faces. Close-up acting is where generated footage breaks — pull out rather than push in. Your practical set covers performance and close-ups; AI covers scale.

4. Let story continuity choose which environment shots to keep. Select generations that match the physical action in your anchor — if your practical still shows someone on hands and knees coming out of water, a shoreline wide cuts cleanly into it. Add environmental continuity cues, like matching haze between the generated exterior and your practical interior, so the transition reads as one location.

5. Lock the look on the first shot, then iterate conversationally. Work with the invideo agent in a back-and-forth dialogue: pin a generation you like and ask for variations — "create another one like it, but change X, Y, and Z." Once the look, tone, and effect are established on the first shot, subsequent shots in the scene come faster and more accurately; in the cave-wall production, the second shot landed correctly on the very first generation.

6. Splice and blend in post. First generations are rarely final — extract the best moments from multiple imperfect clips and cut them together into one sequence. Then apply a unified color grade across practical and generated footage, and add sound design; sound is the multiplier that makes the extended environment feel physically real rather than composited.

Watch some of these to see what works for you:

You definitely get so much more out of AI when you actually use your own footage as the references. You can make it look very close to something that looks real.

— Alex Arfaoui, filmmaker