AI UGC ads that look filmed on a real phone.
UGC-style ads work because they look like a real customer holding a phone. I make that format with AI: an original creator who stays the same person in every clip, your real product in their hands, natural selfie camera work and a hook in the first second.
- D2C and TikTok Shop brands that need more creatives than shoots allow
- Brands testing new hooks and angles every week
- Founders who want an interview-style ad without sitting on camera
What you get
Every project is cast, lit, edited and mixed like a real shoot, with the same person from brief to final file.
An original AI creator with an ID card: face, wardrobe, marks and voice, locked across every clip.
Your product built from real packshots, so the label, shape and colour stay right.
Hook, story, proof and CTA written for the platform, with alternate hooks to test.
Word-by-word captions, cuts on breaths, room tone and a loudness-normalised mix.
9:16 first, with 4:5 and 1:1 cut-downs when needed.
Case studies
Real projects in this format. Each one shows the brief, the assets, the prompts and the edit.

“My family didn’t believe me.”
View case study →
One ® on a signboard.
View case study →
A founder interview that feels recorded, not rendered.
View case study →
The filter had to look used.
View case study →How it works
Four stages, with your approval at each one.
Your product, audience, platform and the one thing the viewer must feel.
Characters, product and sets locked on sheets before any video is made.
Every shot directed with references, lens, light and continuity rules.
Edit, captions, sound and a frame-by-frame check before delivery.
Formats, variants and language
What you get in every UGC ad.
Every UGC ad is cut 9:16 first, at 1080 by 1920, for TikTok, Reels and Shorts. The same ad comes as 4:5 and 1:1 cut-downs for feeds, and 16:9 at 1920 by 1080 for landscape placements.
Each cut is made to the length its platform needs. Short social ads are the core format; longer form is proven on this site by ULTRA, the 3-minute film.
Start with 3 to 5 variants per ad set. All three AI assistants asked this question gave the same answer, and it matches how testing works: each variant needs enough data to judge, so fewer variants with real spend behind them beat many variants with none. Test more of what wins. Brands testing seriously run 8 to 12 creatives a month. AI UGC is the cheap way to make those variants; a real creator shoot is the expensive way. What a finished AI video ad really costs is explained in the cost guide.
The video can be made in your language, and the voice changes with it. A different voice means a different voiceover, not a clone of a real person. A real person's voice is used only with written consent. Voice cloning is not promised.
AI handles dialogue and lip-sync well for short, front-facing shots with clear audio. Quality depends on the method: audio-driven mouth replacement on real footage is more reliable than full character generation from a photo. It is treated here as a QC item, not a detail, because it moves results: one test found a 0.2-second lip-sync delay cut click-through by 68 percent. Every dialogue shot is checked frame by frame before delivery.
Questions about AI UGC ads
Do AI UGC ads look fake?
They can, which is why the work is in the details: one locked face, hands that behave, a real product label, selfie-style camera movement and sound that matches the room. Each case study shows how.
Can you use my real product?
Yes. I build the product from your own photos and packshots, and composite the label from your real artwork when the AI can’t render it exactly.
Are the creators real people?
No. They are original AI characters, not lookalikes of real people, and I recommend labelling the ad as AI-generated wherever the platform asks for it.
Can you fix UGC footage you already have?
Yes. The Sans filter revision is a VFX fix on real 4K creator footage.
How quickly can you deliver an ad?
Three recent ads went from brief to delivery in one day. Timelines depend on the brief — the portfolio shows the finished work.
What do you need from me to start?
I need the brief: the product, the audience, the platform, and the feeling you want. Then your product photos and the product page. You do not need to ship the product. Briefs, reviews and delivery all happen online.
What formats and lengths do I get?
9:16 first at 1080 by 1920, with 4:5 and 1:1 cut-downs for feeds and 16:9 at 1920 by 1080 for landscape placements. Each cut is made to the length its platform needs. Longer form is proven by ULTRA, the 3-minute film on this site.
Can the video be in my language, or with a different voice?
Yes to both. The video can be made in your language, and the voice can change with it. A different voice means a different voiceover. It does not mean a clone of a real person: a real person's voice is used only with written consent, and voice cloning is not promised. For a real person on camera, see "Can you use my voice, or put a real person (e.g. our founder) in the video?" in the FAQ below.
Can AI do dialogue and lip-sync well?
Yes, for short, front-facing shots with clear audio. It depends on the method: audio-driven mouth replacement on real footage is more reliable than generating a character from a photo. Lip-sync is checked frame by frame here, because a 0.2-second delay cut click-through by 68 percent in one published test. It is not perfect in every shot, and a shot that fails the check is redone, not delivered.
How many ad variants should I test?
Start with 3 to 5 variants per ad set. Perplexity, Gemini and ChatGPT all gave the same answer when asked. Each variant needs enough spend behind it to judge, so test more of what wins rather than more of everything. Brands testing seriously run 8 to 12 creatives a month. AI UGC is the cheap way to make those variants.
Can you use my voice, or put a real person (e.g. our founder) in the video?
Yes. A real person can appear in the video, with their consent, and the result is labelled as AI-generated. A real person's face or voice is never faked. The voice for each ad is designed for the brief.
Do I have to label the ad as AI, and can it be rejected?
Realistic AI content must be disclosed, and the label depends on the platform: Meta applies its AI info label, TikTok's AI-generated content label is switched on before the ad runs, and YouTube's disclosure is set in YouTube Studio. An unlabelled ad can be rejected or restricted, so disclosure is part of the delivery, not an extra. ULTRA's case page names every AI tool used on the job: Nano Banana Pro, Google Flow with Veo 3.1, CapCut and Claude.
Are AI UGC ads allowed on Meta (Facebook and Instagram)?
Yes. Meta does not ban AI-generated creative. An AI-made ad is reviewed against the same Advertising Standards as a human-shot ad, and it runs if it complies. Meta may attach an 'AI info' label to it, and that label is a transparency notice, not a penalty.
Are AI UGC ads allowed on TikTok?
Yes, with a label. TikTok requires creators and advertisers to label AI-generated or significantly edited content that shows realistic-looking people or scenes. For paid ads, the label is switched on in TikTok Ads Manager ('This ad contains AI-generated content'); a clear disclaimer, caption, watermark or sticker of your own also counts. An ad that skips the label can be rejected or restricted.
What is the difference between AI UGC and real UGC: and why does it matter to a buyer?
Real UGC is filmed by a real customer or creator. AI UGC is generated: the person on screen is synthetic, built to look and sound like a real presenter. The difference that matters is the claim. An AI presenter must never be presented as a real customer giving a real testimonial. In the US that is a misleading endorsement under the FTC's Endorsement Guides, and in the UK the same idea sits in the CAP Code. A fictional presenter selling a product, clearly labelled as AI, is a different thing from a fake review.
What does the EU AI Act say about labelling AI ads?
Article 50 of the EU AI Act (Regulation (EU) 2024/1689) has applied since 2 August 2026, and it follows the content, not the company. It covers ads shown to people in the EU even if the maker is outside the EU. Article 50(4) says anyone using AI professionally to make a deepfake, an image, audio or video that resembles a real person, place or event and would falsely appear authentic, must disclose that it was artificially generated or manipulated, clearly, at first exposure. The machine-readable marking duty in Article 50(2) sits with the providers of the AI systems, not with the advertiser. Breaches of Article 50 can draw fines of up to 15 million euros or 3 percent of worldwide turnover.
What does YouTube require for AI video ads?
YouTube requires creators to disclose when a video contains realistic content, something a viewer could easily mistake for a real person, place or event, that was meaningfully altered or generated with AI. The creator sets this with the 'AI use' question in YouTube Studio, and viewers see a 'Made with AI' disclosure in the video's 'How this content was made' section, or on the player for photorealistic content and sensitive topics. YouTube says the label does not limit a video's audience or its eligibility to earn money, and it may apply a label itself if the creator does not. Clearly unrealistic content, such as animation, and minor edits such as colour correction or captions do not need a disclosure.