How do you maintain consistent lighting across multiple AI-generated video shots?

Lighting stays consistent across AI video shots when you lock it once at the project level instead of re-describing it per clip. Four methods work:

1. Lighting rules in a persistent style document
2. A fixed lighting-source slot in every prompt
3. A locked style block repeated verbatim
4. Reference-to-video chaining between clips

invideo is an agentic video creation tool that holds your project's style context persistently across every generation, which is what makes project-level lighting lock possible.

Encode lighting rules in a persistent style document. Write your lighting grammar as explicit rules — light sources, shadow color, contrast ratios — in a visual-language document and load it into the invideo agent once at project start, so the invideo agent checks every generated frame against those rules instead of you re-prompting lighting per shot. One documented production stored a director's signature 85:15 dark-to-light ratio this way, and the invideo agent caught shadows "leaning blue-green instead of neutral gray" mid-generation, pulled the relevant rule from the document, and offered a warmer pass without being asked. Once the document is loaded, a three-word continuation prompt — "Everything should match" — is enough to hold lighting, lens grammar, and spatial logic across a multi-shot sequence.

Reserve a fixed lighting-source slot in every prompt. Use a fixed prompt assembly order so lighting is never an optional descriptor that drops out: one documented production held a 9-element sequence (camera spec, lens and aspect ratio, lighting source, palette, composition, atmosphere, mood register, film attribution, negative prompt) across every frame. Write the slot in reference-specific language — "warm yellow from the lamps only, like all the refs" produces more accurate results than a generic "warm lighting" — because color-temperature drift is the most common way lighting consistency fails.

Repeat a locked style block verbatim in every prompt. Upload a batch of style-reference frames — one production used 64 frames from its target aesthetic — and instruct the invideo agent to deeply understand the style and save it to context for all further generations. Include explicit prohibitions in the block ("not live action, not photorealistic") to prevent drift, then start every single prompt with the block; that discipline held one look, including light quality, across 164 generated clips for a 3-minute episode.

Chain clips through reference-to-video. For continuity between adjacent shots, clip the final seconds of each generated segment and re-upload it to the invideo agent, which attaches it to Seedance 2.0 reference-to-video alongside your character and location references. Because the model ingests the full prior video rather than a single frame, it carries lighting, atmosphere, and camera movement across the boundary — which is why it holds light continuity better than start/end-frame or extend methods that see only one frame of context. Every current video model runs inside invideo, so the invideo agent handles routing per shot without you switching tools.

Two adjacent moves help at the edges: use your locked world and environment images — not loose mood-board references — as generation seeds so new shots inherit the established light, and apply one unified grade (slight blur, grain, color matching) across all clips in post to even out residual temperature drift.

These are some of the ways to problem-solve this — what works depends on your shot.

Watch some of these to see what works for you:

I was generating Scene 1 and before I noticed anything, the agent caught that the shadows were leaning blue-green instead of neutral gray. Pulled the Stage A rule from the doc, flagged the deviation, offered a warmer pass. I never asked it to cross-check.

— invideo's creative team, documenting a production run on the invideo agent