If I give you one brief + a few prompts, can you turn it into a complete 9:16 AI ad directly with Codex?
In this tutorial, I’ll share the exact Flova Skill + Codex workflow I use — from brainstorming and writing a 30-second script, to character consistency, storyboarding, and final video generation — using 5 copy-and-paste prompts.
You don’t need to figure out prompt writing from scratch. Just follow this workflow:
Brief → Idea → Script → Character → Storyboard → 9:16 Video
and build the full ad step by step.
What You’ll Get
You’ll learn:
- 5 prompts you can copy directly into Codex / Flova
- How to use the Reference-to-Video Skill properly
- How to learn pacing and structure from a reference without copying it
- How to keep characters consistent across different shots
- How to approve the script and storyboard first, so you don’t waste credits on unnecessary video generations
Let’s start with the first workflow.
What You Need Before You Start
Prepare a reference video you have the right to use for pacing and shot-language analysis, product or SKU assets, a character avatar image if needed, and a short brief covering the audience, product claim, CTA, aspect ratio, and target duration.
Demo 1: Learn from a Reference Without Copying It
Upload the reference to Asset Library, choose a reference-to-video Skill, and add your product assets. The prompt below tells Flova to study pacing, deadpan humour, and shot language only—not to reuse footage, brand elements, characters, copy, or logos.
Use this reference ONLY to learn its pacing, deadpan absurd humour and shot language. Do NOT reuse its footage, brand, characters or copy. Create my own original 30-second version with Traditional Chinese subtitles.
[Paste your product description here]
Once the storyboard is approved, use:
Generate all shots and assemble on the timeline.
Copyright reminder: a reference should inspire structure and energy, not copied footage, branding, characters, copy, or logos.
Demo 2: Create an Ad from Codex with Flova CLI
Open Flova’s CLI page and paste this into Codex:
Please help me install Flova CLI: https://cli.flova.ai/flovaCLI-setup.md
If browser authentication fails, restart Codex and try flova auth login or flova update. If necessary, generate an Agent API key from the CLI page, then run this in PowerShell:
flova auth login --token "YOUR_FLOVA_API_KEY"
Treat the API key like a password. Never place it in public chats, source code, screenshots, or logs. Then verify the connection:
Read my Flova project status.
Five Prompts: From Idea to Final Cut
1. Brainstorm five concepts
I'm making a viral ad for a canned energy drink. Brainstorm 5 exaggerated, absurd, highly shareable concepts. Use the tone of over-the-top deadpan "big game" commercials as inspiration only — original ideas, no existing brands. Give me a one-line logline for each.
2. Write a 30-second script
Take concept #3. Write a 30-second ad script: a strong 3-second hook, absurd escalating humour, a clear product hero moment, and a punchy CTA. Keep it PG.
3. Create a three-view character sheet
Use Flova to turn the attached avatar into a 3-view character sheet (front, side, back) for consistency across all shots.
4. Build the storyboard
Storyboard the 30-second script using this character sheet. Keep the same character consistent across every shot.
5. Generate and export
Generate all shots, then add voiceover and music, and export a 9:16 cut.
How to Spend Fewer Credits
Do not generate video as soon as you have an idea. Refine the script and storyboard first. Images are generally cheaper to iterate on, while video is commonly charged by duration. Review longer shots, faces, and on-screen text manually before publishing.
The key idea is simple: your AI agent does the thinking; Flova makes the video. Once you save your product assets, character sheet, and successful workflow, you can create campaign variations much faster.
