π₯Έ Mustachio AI Video Ad
Made it with a triple combo β ChatGPT 5.6 Sol + Seedream 5.0 Pro on @higgsfield_ai + Seedance with @capcutapp .
this method is faster in development, but you trade away a bit of control.
βΎ storyboard and prompt below
Prompt itself is short. It's working good only with visual storyboard reference.
Create a polished 15-second photorealistic 1980s American moustache-cream commercial.
Follow the reference storyboard exactly from the first frame to the last: use the same shot order, timing, compositions, character performance, product, moustache transformation and final typography. Treat the reference only as a storyboard guide; never show the grid, panels, captions or timecodes in the video.
Add beautiful background music inspired by elegant late-1970s and early-1980s American commercials.
Warm, expressive American male voice-over:
βHe was ready for the day. His moustache wasnβt. Then came Mustachio. Make your moustache happy.β
Synchronize the final spoken slogan with the typography appearing word by word.
Tested ChatGPT Sol for storyboard prompts. fed it my moustache cream concept, passed the result to Seedream 5.0 Pro on @higgsfield_ai
βΎ full prompt below
STYLE
Photorealistic cinematic advertising storyboard inspired by authentic 1980s American menβs grooming commercials. Warm 35mm film texture, soft natural grain, slightly faded color response, clean realistic skin and hair, understated visual comedy.
FORMAT
One 16:9 storyboard sheet divided into six equal cinematic panels in a clean 3Γ2 grid. Consistent character, bathroom, wardrobe, lighting and product design across every panel. Each panel has a small black caption strip with exact white text.
CHARACTER
A handsome 35-year-old American man with wavy dark-brown hair, a thick natural dark-brown moustache, expressive eyebrows and realistic facial texture. He wears a plain white ribbed tank top.
LOCATION
A stylish middle-class American bathroom from the 1980s: pale mint ceramic tiles, cream sink, chrome fixtures, glass shelf, warm morning sunlight and soft tungsten practical light.
PANELS
Panel 1 β Extreme close-up of the lower face. The man smiles, but both moustache tips droop downward like a sad frown. Caption: "SHOT 01 β SAD MOUSTACHE" and "00:00β00:02".
Panel 2 β Medium mirror shot. He smiles wider while the moustache remains visibly sad and drooping. His eyes look down toward it with mild confusion. Caption: "SHOT 02 β SOMETHING IS WRONG" and "00:02β00:04".
Panel 3 β Close product insert <<<image_1>>>. A stylish round jar of moustache cream <<<image_1>>> labeled "MUSTACHIO" and "MOUSTACHE CREAM" stands on the glass shelf as his hand reaches toward it. Caption: "SHOT 03 β MUSTACHIO" and "00:04β00:06".
Panel 4 β Tight facial close-up. He applies a small amount of cream through the moustache with one simple outward finger movement. Caption: "SHOT 04 β APPLY" and "00:06β00:08".
Panel 5 β Frontal close-up. The moustache becomes neat and full; both tips lift and curl upward like a smile. The man smiles with satisfied surprise. Caption: "SHOT 05 β HAPPY MOUSTACHE" and "00:08β00:12".
Panel 6 β Clean packshot of the same <<<image_1>>> "MUSTACHIO" jar centered on the glass bathroom shelf, softly blurred mint-tile background, warm sunlight, premium 1980s advertising composition. Text beside the product: "MAKE YOUR MOUSTACHE HAPPY." Caption: "SHOT 06 β PACKSHOT" and "00:12β00:15".
CONTINUITY
The moustache is naturally drooping in panels 1β4 and symmetrically lifted in panel 5. Preserve the exact same man, jar, label design, bathroom and morning lighting throughout. All typography must be clean, readable and limited to the exact quoted text.
@JSFILMZ0412@dreamina_ai And again - it's looks perfect, if you are not using v2, wow, it's real challenge. wanna do something similar. inspired me to grow up in music videos, thank youβοΈ
Two weeks of active work with ChatGPT 5.6 Sol and Claude Fable 5. Both are good, but Iβm staying with ChatGPT.
It all comes down to limits.
X has been full of Claude hype lately, and itβs not coming from nowhere. The models are genuinely good. ChatGPT released its answer a little later, and after working with both, I can get the result I need from either one.
What surprises me is how people compare which LLM writes better prompts for image and video generation.
Wait. I write the prompts, not the LLM.
I control the context, language, materials, scene, and camera. Give either model too much freedom and both will eventually write nonsense or lose context.
Itβs still human in the loop. The model helps, but the control stays with me.
What I really discovered after trying Fable, Opus, and even Sonnet was the limits.
I work with LLMs a lot every day. Even small requests drain the five-hour limit, and the weekly allowance can disappear in just a couple of days.
ChatGPT Sol, meanwhile, gives me agent mode, the same kind of skills for deep specialization, and native image generation. I use all of it heavily and have never hit a limit. I only found the limits menu for the first time after my experience with Claude.
So I wonβt renew my subscription.
Not because the model is bad. I just donβt see the point. Both give me the result I need, but only one lets me work without constantly thinking about when the limit will reset.
For me, the choice is obvious. ChatGPT simply offers better economics.
Nothing personal. Thanks to healthy competition.
How is it on your side? Feel free to leave your experience in replies.
@capcutapp@CinaimaFilms It's start from prompt to agent and no choose to generate few images or video iteractions. can't exclude this agent from UI. just wanna button without ai chat... CapCut team. it's a basic, standard in 2026.
@CaliraVal@wavespeed_ai In this ser seedance is lost control... On storyboard one look, at the video another one. I think it's can be fixed better in the prompt. Anyway - looks cool βοΈ
@Shorelyn_@ImagineArt_X Great set! Bravo. But who yawned that loudlyβher boyfriend? π I hope Seedance eventually lets us export the music and sound effects as separate audio tracks so we can edit them.
@_OAK200 It's look cool in aesthetic but as I see on the wide plan - man have 7 fingers? Waiting Mj fix that. I think normal anatomy is the standard at 2026...
Hey! Prompting an action scene isnβt that difficult. But getting an interesting anime style is another story. Could you show or explain which reference images you usedβor perhaps share an image along with the style description? Iβd be really interested, because as you said, getting the model to produce anime without defaulting to something like the Ghibli style is a real challenge.
Also - cool video man!