Keep going
Every workflow on Rundown University, free for 7 days
Members get all guides, courses, live sessions, and $1,000+ in partner perks.
Guides Higgsfield AI published feb 12, 2026
In this guide, you’ll learn how to create a professional-looking TV commercial using AI. The same workflow can also be used for YouTube ads, Facebook ads, or other short video ads. The source workflow uses $9 software and takes around 20 minutes.
You’ll build a short AI-generated commercial from a two-scene script, four generated key frames, and two AI-generated video clips that are spliced into a final ad.
Use photography terms to describe the camera and angles of your frames, such as anamorphic lens, macro lens, 14mm wide angle, hero shot, or Dutch angle.
Go to Gemini.google.com and describe your TV commercial idea. Ask Gemini for a 20-second script with exactly two scenes. Review the script and revise it as needed.
If you already have a custom voiceover, sound effects, or music in mind, include those details in the script request.
I have a TV commercial idea: [describe your idea]. Write a 20-second script with exactly two scenes. Include voiceover, sound effects, or music notes if relevant.Next, ask Gemini to generate prompts for the start frame and end frame of each scene. Tell it to use photography and film terms to keep a consistent aesthetic.
Generate prompts for the start frame and end frame of each scene. Use photography and film terms to keep a consistent aesthetic.💡 Pro tip: Use photography terms to describe the camera and the angles of your frames. Try phrases like: anamorphic lens, macro lens, 14mm wide angle, hero shot, or Dutch angle.
Sign in to Higgsfield.ai. Click Image > Create Image > Select Nano Banana Pro.
Set the quality to 4K and variations to 4.
Set the aspect ratio to 21:9. This is the cinematic aspect ratio Google uses when training Nano Banana on films.
Paste in the prompt for your scene 1 start frame. When it is finished, choose the version you like best and click the heart button so you can find it later.
Repeat the process with the scene 1 end frame prompt and your scene 2 frame prompts. By the end, you should have 4 images total.
Download the 4 images to your desktop.
💡 Pro tip: Drag the start frame into your prompt box along with your end frame prompt. This helps keep characters and settings consistent within each scene.
In Higgsfield, navigate to Video > Create Video and select the latest Kling model. The source example uses Kling 3.0.
Upload your start frame and end frame images.
Copy this scene’s section from your Gemini script and put it in the prompt box. Adjust the clip length as desired, then click generate.
Repeat the same process for scene 2.
When both scenes are generated, download them from your asset library in Higgsfield.
Splice the scenes together in a free desktop app like CapCut. Alternatively, you can ask Claude Code, Codex, or Claude Cowork to splice them together using ffmpeg in the terminal.
For next-level polish, focus on audio. Try muting your generated clips, then use CapCut’s audio library to add sound effects and music.
You can make even better audio using tools like Suno for music and ElevenLabs for voiceovers.
Keep going
Members get all guides, courses, live sessions, and $1,000+ in partner perks.