What's the best way to handle props or handheld objects when setting up a character for turnaround sheet generation, and are there any tools or workflow steps that help ensure the character is properly prepped before running it through the process?

Remove handheld objects from the character before generating a turnaround sheet. Generate the character empty-handed in a neutral pose, and generate each prop as its own separate reference asset. A held object forces the model to keep its geometry accurate from every angle simultaneously — hand-object contact points are where multi-angle sheets break.

Generate the character clean and the prop on its own sheet — that single rule prevents most turnaround inconsistency. A turnaround asks the model to render the same forms correctly from four or more angles at once, and physical contact between hands and objects is the weakest link: the same failure mode breaks AI video models faster than anything else when props, ropes, and bodies touch. A sheet error also doesn't stay in the sheet — the turnaround becomes the source of truth attached to every downstream generation, so a malformed prop baked into one panel propagates through the entire film. Removing the object before generation beats inpainting hands panel-by-panel afterward, because the sheet is what every future shot reads.

Give the prop its own reference pass. One documented production generated 11 reference images total — headshots and head-to-toe references for 4 characters plus a separate sheet for 1 prop, a necklace. Direct the prop like a narrative object: generate multiple alternatives and select on story logic, not just looks, and put physical characteristics in the brief — one production specified "hard material, so it makes a horrible sound when it falls" so the object carried sound logic into later shots. If the final film genuinely needs a fused character-plus-object sheet, a quick hand-drawn sketch of the arrangement uploaded as a reference image can anchor what text prompts alone won't produce.

Prep workflow before running the sheet — invideo is an agentic video creation tool with the current image and video models available, and the invideo agent runs each of these steps:

1. Answer the spec questions first. The invideo agent surfaces pre-production questions before building any asset — character description, entity reference, prop specification, deliverable format — because those answers change every frame. Lock the prop spec on paper before any pixels.
2. Go frames-first: portrait, then turnaround. Lock the face with a portrait pass before generating the full sheet; the invideo agent routes each asset to the right image model. In one documented workflow the sheets ran at 4K with four angles plus face and mid-angle close-ups.
3. Generate options, then lock. Generate four variations of each character sheet, pick the best, and lock it before any video generation — locking assets upfront is the step that prevents consistency problems for the rest of the film. Budget for iteration: locking one character ran about 5 generations at ~$9.78 per character in a documented production.
4. Include close-up panels. Sheets need face and detail close-ups, not just wide angles — small elements like scars and accessories drift across models without them, and the model hallucinates anything it can't see.
5. Make per-beat sheets if held objects change. When a character picks up or accumulates objects across the story, build a separate sheet for each appearance state — one production needed a distinct sheet per sequence because the character added a trinket in every new location.
6. Fix errors at the sheet, not the shot. If a prop or accessory error surfaces mid-production, ask the invideo agent to inspect the sheet rather than re-rolling shots: in one documented case it identified the exact panel containing a stray earbud, corrected it, and stored the updated sheet in context so every subsequent shot inherited the fix.

This prep is what makes sheet-based consistency hold without fine-tuning: one 70-second production kept 2 characters identical across every scene using locked character sheets and the invideo agent's context system — no LoRA required.

Watch some of these to see what works for you:

the AI always needs to see what the character is exactly, right? Or else it'll kind of hallucinate and imagine something that's under the cap. So, we don't want to do that. We always want the character to be seen as we see it on the character sheet.

— invideo's creative team