The brief was simple and a little uncomfortable. Could we produce a UGC-style beauty video using AI tools that a real person scrolling Instagram would not immediately identify as AI-generated?
In early 2026, our creative team ran this as an internal production experiment, working with Lakmé’s publicly available brand language as the visual and tonal reference. The goal was not to create a polished brand film. It was to build something that looked like a real creator had filmed it in their bathroom, talked naturally about a product they genuinely liked, and uploaded it without a production team involved.
We documented the entire process. This is what we learned.
The Brief
UGC (User-Generated Content) video has become one of the highest-performing ad formats in beauty and personal care on Meta and Instagram. The format works because it signals authenticity. A real person, in a real setting, speaking naturally about a product they actually use performs better than the aspirational polish of traditional brand advertising.
The challenge for brands is that real UGC is unpredictable. Quality varies. Brand guidelines are hard to enforce. Scaling UGC content production means managing many creators simultaneously.
Our brief for this project had four success criteria. The video needed to match the visual and tonal quality of organic creator content. It needed to be produced significantly faster than a traditional shoot. It needed to be indistinguishable from genuine UGC in a social feed context. And it needed to be adaptable for A/B testing without reshoots.
The AI Tool Stack
We used a combination of generative AI tools, each handling a specific element of the video.
For script and narrative, we used Claude for script development and voiceover copy, guided by Lakmé’s established tone of voice and product claims. For visual generation, we used Midjourney for still image reference generation, and Runway Gen-2 and Kling AI for video clip generation from text prompts and reference images. For audio, we used ElevenLabs for natural-sounding voiceover in a conversational register, slightly intimate rather than broadcast-polished. For edit and assembly, we used Adobe Premiere Pro for final assembly, colour grading, and caption integration.
The Production Process
Phase 1: Visual Direction (2 days)
Before generating any video, the team developed a detailed visual brief. Lighting style (bathroom counter, warm ambient, no ring light effect), colour palette (consistent with Lakmé’s warm skin-tone positioning), presenter characteristics (South Asian, mid-20s, casual but composed), and the specific gestures typical of high-performing beauty UGC, such as holding the product up, applying to the back of the hand, and a natural finishing moment.
Reference gathering from organic TikTok and Instagram UGC in the beauty category informed both the visual direction and the script structure. The best-performing UGC formats in beauty follow a consistent structure: a hook presenting a relatable pain point or transformation claim, a demonstration of product application or result, social proof that builds trust, and a CTA that feels natural rather than scripted.
Phase 2: Script Development (1 day)
The script was written to sound spoken, not read. Beauty UGC that converts sounds like someone talking to a friend. This means incomplete sentences, natural pauses, conversational transitions such as “and honestly” and “I was not expecting this”, and a structure that follows curiosity rather than a formal narrative arc.
We tested the script for readability by reading it aloud three times. Any phrase that felt unnatural when spoken was rewritten.
Phase 3: Visual Generation (3 days)
This was the most iterative phase of production. We generated 4 to 6 variations of each key visual moment, selected the best performer for consistency and quality, and used that clip as a visual reference for subsequent generation to maintain character continuity.
Approximately 40 clips were generated to produce the 12 usable clips in the final video. A 3-to-1 generation-to-use ratio is a realistic benchmark for current AI video production at quality levels appropriate for paid social.
Phase 4: Audio Production (1 day)
The voiceover was generated using ElevenLabs with a voice profile selected for natural warmth and credibility. Key parameters: slight variation in pace, natural breathing pauses, and delivery that matches the conversational script register. Multiple takes were generated and the most natural-feeling delivery was selected for each segment.
Phase 5: Assembly and Post-Production (1.5 days)
Final assembly in Premiere Pro combined the selected clips with the voiceover, text overlays (captions are table stakes for Reels and TikTok, with most views consumed without sound), subtle colour grading to maintain consistency across clips from different generation sessions, and ambient audio to reduce the sterility of a completely clean voiceover track.
Total production time: 8.5 working days from brief to final deliverable.
Results and Assessment
The video was used as a paid creative test on Meta, A/B tested against a traditionally-produced UGC-style video from a real creator.
| Metric | AI-Generated | Traditional UGC |
|---|---|---|
| 3-second view rate | 42% | 39% |
| Full video completion | 28% | 31% |
| CTR to product page | 1.8% | 2.1% |
| Cost per click | ₹14 | ₹16 |
| Production time | 8.5 days | 12 days |
| Production cost | ₹45,000 | ₹85,000 |
The traditional UGC slightly outperformed on completion rate and CTR. The AI-generated version had a clear advantage on production time, production cost, and adaptability. We reskinned the AI version for a different product variant in 2 days. A reshoot of the traditional version would have taken a minimum of 10 days including creator coordination.
What This Means for Brands
The honest conclusion from this case study is that AI video production in 2026 is not a wholesale replacement for high-quality traditional production. It is a powerful complement that changes the economics of content production.
For categories where volume matters more than perfection, such as social media content, A/B test creative, and localised adaptations, AI video production delivers significant efficiency gains.
Our AI Video Production service is built on this hybrid philosophy. Explore our broader AI Marketing Systems or contact our team to discuss what AI production could look like for your brand.
Frequently Asked Questions
Can AI-generated video pass as authentic UGC?
In a social feed context, well-produced AI UGC video is frequently indistinguishable from human-created content. The variables that matter most are script naturalness, character consistency, and audio quality.
How much does AI video production cost in India?
A complete AI UGC video production, from brief to final deliverable as in this case study, typically costs ₹35,000 to ₹75,000 depending on length, complexity, and number of revisions. This compares favourably to equivalent traditional production at ₹75,000 to ₹2,00,000.
Is AI video suitable for performance marketing creative?
Yes, particularly for social media ads where creative refresh rate matters. The ability to produce and test multiple creative variants quickly makes AI video production well-suited to performance marketing workflows where A/B testing is standard practice.
