AI Open Guide
← All goals

Short clips, ads and social videos, with the same character every time.

Best tools for this

HiggsfieldMany video models, effects and a character tool.SeedanceRealistic clips with sound, up to 30 seconds.GeminiShort clips inside the Gemini app.

Read these chapters

Your first video
Higgsfield · Chapter 1.4 · 3 min readHiggsfieldYour first clip
Seedance · Chapter 1.4 · 3 min readSeedanceA good video prompt, block by block
Higgsfield · Chapter 1.8 · 3 min readHiggsfieldHow a Seedance prompt is written
Seedance · Chapter 1.5 · 3 min readSeedanceCharacters that stay the same (AI Influencer)
Higgsfield · Chapter 1.9 · 3 min readHiggsfieldMaking short videos (Veo)
Gemini · Chapter 1.10 · 3 min readGemini

Ready prompts to copy

Create#1

Make a short 3D product video (plan, prompts, generation)

Goal. A 20–30 second video that shows my product turning, lit like a commercial, ready for social media.

Open. Cowork, Opus or Fable model, effort High; a video-generation connector enabled (for example Higgsfield) with credits; attach a clean product photo. *Needs a tool outside Claude.*

You are a cinematic director and AI video prompt engineer.
Goal: a 25-second 3D reveal of [product], for [platform], [9:16 or 16:9].
Context: audience [who]; brand colors [colors]; mood [calm / energetic / luxurious]; no humor.
Inputs: attached product photo; brand guide if any.
Constraints: 4 to 6 clips of 4–8 s; photoreal; no on-screen text; [music or silent].
Process: ask me up to 3 questions if anything blocks you; then write a shot list; then write one generation prompt per shot (subject, action, camera, lens feel, light, mood, what to avoid); review your prompts as a harsh director; then generate the clips with the connector and show me each.
Output: a table (shot, duration, camera move, prompt), then the generated clips.
Quality bar: a stranger could paste each prompt into a video model with no edits.
What you get · Watch out

What you get. A shot table, then 4–6 clips you can download and join in any editor.

Watch out. A blurry product photo gives a blurry product; use a clean one on a plain background. Generation costs credits per clip; generate one shot first and approve it before the rest. If stuck: Chapters 2.1, 4.2, 4.5.

Create#21

Goal. One invented presenter who looks the same in every video.

Open. Cowork with a character/video tool connected (for example Higgsfield's character features); keep a "character bible" document in a Project. *Needs a tool outside Claude.*

Help me create a fictional presenter for [channel / purpose]. Audience: [who]. Personality: [3 traits]. Look: [age range, style, setting]. Never a real person's likeness.
1. Write a character bible: name, backstory in 5 lines, voice and catchphrases, what they never say, visual description precise enough to repeat (face, hair, clothes, colors, lighting).
2. Write the reference-image prompt for the tool, then a prompt template for scenes that always restates the look.
3. Generate the reference images with the connector and show me; iterate until I approve one.
Save the bible as a document I can add to my Project.
What you get · Watch out

What you get. A character bible document, a reusable scene prompt template, approved reference images.

Watch out. Do not base the character on a real person. Put the bible in a Project so every future video starts from it. If stuck: Chapters 1.5, 4.3.

Build your own prompt →