Asset Prompts

GPTClaudeDeepSeek··▲ 1,310 copies·updated 2026-07-14

asset-prompts.prompt

# Asset generation prompts — patterns

Read this at workflow step 2, when you extract assets and assign @names. Every asset in the Asset Checklist gets a **copy-ready English generation prompt** the user pastes into an image tool (Higgsfield Soul / GPT Image / similar) to build the reference BEFORE video generation. The board renders each one as a copyable block (markup in `board-spec.md`).

Source method: Higgsfield Cinema Studio workflow (higgsfield.ai/blog/cinematic_headphones) — assets first, locked as Elements with @names, then video prompts reference them.

## Rules that apply to every asset prompt

- **English, photorealistic, self-contained** — the prompt works standalone in an image generator.
- **No branding**: append to product/wardrobe/vehicle prompts — "STRICTLY no brand names, logos, wordmarks, or text" (exception: the client's own brand mark when the ad requires it).
- **Plain background for characters/props**: "plain solid grey background" — clean references pin identity better.
- **No people in location references** — locations must be empty stages.
- **No readable signage/text** in locations and UI props — text renders badly and pollutes the reference.
- Match the film's palette and time-of-day in location prompts (the reference teaches the model the look).

## Character sheet (the split-frame pattern)

One image, two panels — face for identity, full body for wardrobe and build:

when to use it

Community prompt sourced from the open-source GitHub repo afloy011-spec/seedance-shotlist-director (NOASSERTION). A "Asset Prompts" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.