Goose the Ads Guy

I make 100+ videos every week using Claude Code and I can teach you how.

In my day job, I also help brands scale their Meta and TikTok ads.

[Goose the Ads Guy @GooseworksAI]

@sohmehta @AEO 👀👀

[Goose the Ads Guy @GooseworksAI]

We make ads like these everyday 🪿🪿🪿

[Soham Mehta @sohmehta]

Agencies charge $50k+ and six weeks for a spot like this.

I made it solo, in a terminal, in 2 hours.

No camera. No studio. No model. No editing software. Every frame generated with AI.

A full American Eagle (@AEO) ad 👇

───
Step 1 — the brief.
I typed one sentence: "a girl, denim on denim, the jeans are the hero, sage-green studio, fashion-film energy."
That's the entire creative direction I gave. Claude Code (@claudeai) took it from there.
───
Step 2 — lock the face.
The thing nobody can crack with AI ads: the same woman in every shot. Not her lookalike cousin in shot 4. Her.
CC generated the model and the set and locked them. Identical, start to finish.
───
Step 3 — board it to the beat.
I picked the track first — scored with @suno. Then it storyboarded the entire ad to the song — every cut, every macro, the logo drop — before a single image existed.
The edit was designed before the footage was.
───
Step 4 — keyframes.
Real @AmericanEagle product fed in as reference, so the denim is the actual denim. Every still generated with gpt-image / nano banana on @fal.
That button macro that looks shot on a $20k lens? Generated. Real wash, real stitching, real hardware.
───
Step 5 — make it move.
Then every still starts breathing — the mirror turn, the walk, the hair, the fabric. Each keyframe → video with Seedance on @fal, best take kept.
───
Step 6 — cut it.
ffmpeg stitched every clip into one timeline and snapped each cut to the beat.
No Premiere. No After Effects. A command line.
───
Step 7 — the part nobody posts about.
It watched the cut back frame by frame, hunting its own mistakes — drifting faces, melted hands, gibberish text — and fixed them until they were gone.
This is 90% of why it doesn't read as "AI."
───
Step 8 — finish.
Burned the captions and the lockup, checked the sync, exported.
One sentence in. A finished ad out.
───

This is what we do at @GooseworksAI.

Real ad creative — no agency, no shoot, no six-week wait. In Days, not months, at A fraction of the cost.

Run a consumer brand and tired of $50k invoices and slow timelines? → DM me, or drop a 🪿

[Shiv @shivsakhuja]

It's insane that GPT Image 2 + Seedance 2.0 can do videos like this so reliably.

It's a pretty neat trick: don't just ask for a drone shot.

Draw the exact path you want the camera to follow.

Makes for an awesome opening shot.

[Goose the Ads Guy @GooseworksAI]

Hey I made this

[Soham Mehta @sohmehta]

made @CalvinKlein an ad in Claude Code. one prompt. Goose stars in it.

I'm 20. never used a video editor.

if you run a brand and want ads like this- DM me. first one's free.

@GooseworksAI

[Goose the Ads Guy @GooseworksAI]

@shivsakhuja @higgsfield_ai Hey - I made that one!

[Shiv @shivsakhuja]

You don't need to be paying for SaaS tools to clip your videos.

Claude can watch your videos, find relevant segments, clip and stitch them cheaper, faster and better than most clipping apps. Here's how:

1. Tell Claude Code to download every video from your YouTube library into a folder.

2. Then have it watch all of them and pull out the moments worth posting on short-form.

CC can use a skill /watch (linked in comments) to watch the video and then output the clips into a folder, named so you can actually tell what they are.

The skill uses ffmpeg to watch the video frame-by-frame + transcribes the audio using whisper.

(s/o to bradautomates for this awesome skill)

3. Then ask it to write an index file describing each clip — so next time it knows what's in there and when to use each one.

The index is for CC to be able to re-use the clips next time you need it.

Bookmark this for later. /watch skill is linked below.

[Shiv @shivsakhuja]

You can turn product photos into studio-quality product videos with Claude Code + @higgsfield_ai + the right skills.

Here's the workflow for Claude Code:

0. Install Higgsfield CLI and get a Fal API key

1. Scrape product images from the brand’s website

2. Generate 10 realistic, stylized versions with @openai’s GPT Image 2 or Nano Banana Pro. I use Fal to access these models.

3. Review the generated images, ditch any with hallucinations and keep the good ones.

4. Feed them into @higgsfield Marketing Studio via the CLI and ask for hyper motion videos.

5. Use @suno to make a song.

6. Stitch together with ffmpeg

This is what comes out. No studio, no video editor.

Just Claude Code and a single skill.

Comment skill below and follow – I'll DM you a skill that one-shots videos like this end to end.

[Himanshu Bamoria @0xhbam]

Most AI-generated brand videos still feel generic.

So I tested if Claude Code/Codex + Higgsfield AI could create a brand-grounded content factory.

Turns out, yes.

Here's the workflow I used to create this video for @touchland :

1. Research the brand, products, ICP, positioning, and visual style

2. Turn it into a brand bible

3. Generate video concepts for different customer segments

4. Pull real product images from the website

5. Preprocess them with Nano Banana 2

6. Write a storyboard with consistent scenes/props

7 .Send assets + storyboard to Higgsfield

8. Have Claude self-QC frames and iterate

The bigger opportunity:

Brands can run this across hundreds/thousands of SKUs and generate product videos grounded in real product info, brand assets, and customer segments.

I turned this into a reusable Skill. Let me know if you want access.