Quick Answer
Automating video ads for clients at scale works when you treat it as a modular production problem, not a generation problem. You build a small number of strong creative concepts, then use a system — template-based or AI-generated — to adapt those concepts across products, formats, and copy variations. The quality ceiling is set by the concept. The automation handles the volume.
The Brief That Every Agency Will Eventually Face
A client needs unique video ads for every product in their catalog. Fourteen products. Three formats each. Three copy variations per format. That is 126 videos, minimum.
For a small paid ads team that usually delivers 3–5 hero videos per campaign, this feels like a different kind of project entirely.
It is. But the agencies that handle it well are not working 10x harder. They have changed how they think about production.
Why "126 Unique Videos" Is Usually Not What the Client Means
When a client asks for a unique video for every product, they usually mean they want:
- every product to be represented,
- each ad to feel relevant to the product it promotes,
- enough variation to run a real test across the catalog.
They almost never mean: 126 distinct concepts, 126 different creative directions, 126 entirely separate productions.
The first conversation worth having is clarifying what unique means in the brief. Is it 126 distinct emotional ideas, or is it 126 outputs that correctly feature each product with the right copy? Those are very different problems with very different solutions.
Most of the time it is the second. And the second is very solvable.
The Modular System: Where to Start
The approach that works at this scale is building fewer creative concepts and generating more outputs from each one.
Step 1: Build 3–5 strong creative concepts
Each concept covers a different angle — a different hook, a different emotional driver, a different reason to buy. For example:
- Concept A: social proof / transformation (before and after)
- Concept B: problem framing (this is why you keep failing at X)
- Concept C: product demo (see exactly how it works)
- Concept D: offer lead (limited, urgent, reason to act now)
- Concept E: founder or authority story
These five concepts are your creative work. This is where strategy and craft go in.
Step 2: Build a visual template for each concept
Each concept becomes a visual structure that can accept a product reference and a copy variation. The structure stays the same. The product and the copy change.
This is exactly what template-based tools like Plainly or Bannerbear are built for. You design the template once in After Effects or a browser tool. Then a spreadsheet drives the outputs — one row per product, one output per row.
Step 3: Layer copy variations
For each product × concept combination, write the hook, value, and CTA variations. Three copy sets per product is manageable. You are not writing 126 ads. You are writing 5 concepts × 3 copy variations = 15 text combinations, then applying them across 14 products.
Step 4: Generate and filter
Run the combination. You get your 126 outputs. Not all of them will be equal — some product + concept matches work better than others. Do a fast review pass and filter to the set worth testing. This is still faster than building each video from scratch.
Template-Based vs AI-Generated: Which to Use
Both approaches work at scale. They solve different parts of the problem.
Template-based tools (Plainly, Bannerbear, Placeit)
Best when:
- the client has strict brand guidelines,
- the visual structure needs to be exactly consistent,
- you are working from existing footage or product photography,
- the client needs to approve a master layout before you scale it.
The output is predictable and brand-safe. The tradeoff is that you need a well-designed template to start from, which requires either a designer or good After Effects skills.
AI video generation
Best when:
- you do not have existing product footage,
- the client wants the ads to feel native to TikTok or Meta Reels rather than polished brand video,
- you need to generate lifestyle scenes around a product,
- speed matters more than pixel-perfect brand consistency.
AI video has improved significantly. The output is no longer reliably "AI slop" if you use the right model with a good reference image and a tight creative direction. The agencies getting good results are not prompting blind — they are feeding the model a product image, a scene direction, and a clear brief.
Hybrid: the approach most agencies land on
Use AI generation for the product scene or lifestyle clip. Use a template layer for the text overlays, the brand elements, and the format adaptation. The AI handles the hard part (creating a visual without filming). The template handles the consistency and the copy.
How to Price a 126-Video Project
The pricing instinct for this kind of project is to multiply your usual per-video rate by 126. That is the wrong calculation.
You are not delivering 126 independent productions. You are delivering a system that produces 126 outputs. The creative work is the system — the concepts, the templates, the copy matrix. The outputs are the result of running that system.
A more honest pricing model:
- Creative fee: for developing the 3–5 concepts and the modular structure. Price this like you would a campaign strategy and design retainer.
- Production fee: per output, but at a discounted rate that reflects the repeatable nature of the work. A common approach is to price the first 20 videos at full rate, then discount the remaining volume.
- Setup fee: if the client needs templates built from scratch, that is a separate cost from generating the videos.
One commenter in a thread on this exact scenario put it well: you are not delivering 126x the creative value. You are delivering maybe 3x the creative value with 40x the production capacity. Price for the creative value you are delivering, not the output count.
A practical anchor: if your standard campaign is $5,000 for 5 videos, a 126-video modular project might be $15,000–$25,000 depending on the complexity of the template build and the number of concepts required. Not $126,000.
Quality Control at Scale
The risk with any high-volume automated workflow is that bad outputs ship. A few things that keep quality up:
Review at the concept level, not the output level
Before you generate 126 videos, make sure the template and the concept hold up. Test it with 3–5 products first. If the system produces good results for your sample, scaling it will produce proportionally good results for the rest. If it is weak at 5, it will be weak at 126.
Set a minimum bar and filter fast
Not every output needs to be reviewed frame by frame. Set a clear bar — does the product appear correctly, does the hook text display as intended, is the CTA visible — and batch-review against those criteria. Move fast. The bar is "good enough to test," not "ready for a brand film reel."
Flag product + concept mismatches early
Some products do not work with every concept. A social proof hook works differently for a physical product than for a SaaS tool. Walk through the product list against each concept before you generate and remove combinations that clearly do not fit. You will reduce the output count and improve the overall quality of what you actually ship.
What Clients Actually Want From This
The agency owner in the original thread identified something important: landing this client matters more than the money. The output of a 126-video project is not really 126 videos. It is proof that your agency can handle production at a scale most agencies cannot.
If you deliver it well — organized, on brief, fast — it changes what kinds of clients and projects you get pitched in the future. That is the actual value of building this capability.
The way to deliver it well is not to work harder on 126 individual videos. It is to build a system once and run it. Every agency that figures this out ends up using it on the next project, and the one after that. The investment in the workflow pays back many times over.
---
The agencies doing this at volume today are not bigger or better staffed than the ones still doing it manually. They built the modular system once, proved it works with a willing client, and kept running it. The first project is the hardest. The tenth is nearly automatic.