Automate the transformation. Never automate the invention

A transformation step takes something you produced and changes its form. Speech becomes text. A long take becomes a short cut. Words become timed captions. A cut becomes an MP4.

Those are dependable because the truth arrived with the input. The step cannot invent a claim about your product. It was never asked to produce a claim.

An invention step produces material that did not exist. A script. A testimonial. A product shot. A verdict on whether the ad worked.

Those are unreliable in a specific way. They fail where a viewer needed to trust something specific. They fail while looking competent, which is the expensive combination.

Apply the rule to any feature you are considering. If it transforms your material, the risk is low and the saving is real. If it invents, you are outsourcing the part that decides performance.

Models are good at plausible objects. They are bad at consistent specific ones. Packaging shifts colour between shots. The logo becomes a smudge that is nearly yours.

For anything a customer will recognise, that is not a cosmetic problem. It is the moment a viewer decides the whole ad is synthetic and stops believing the claim as well.

Synthetic performance is the second failure. A generated presenter reading an informational script is fine. The same presenter delivering first-person testimony reads false to most viewers.

Generated scripts are the third. Competent, well-structured and entirely generic. A model trained on the average of the internet returns the average of the internet.

What it cannot supply is the sentence a customer used last Tuesday, why the second version changed, or the return rate. Those are the details that make a claim survive a filter.

The assembly line is the one with a fixed rate available. Everything else in the cost stack is negotiated or unbilled.

In Cutroom a batch is 100 credits. That covers transcription, the director pass and the footage search. An export is 20 credits per output minute, so a thirty second ad is 110 credits.

Basic is $39.99 a month for 2,500 credits. Premium is $79.99 for 5,000. A 1,000 credit top-up is $15 when a month runs long.

Those numbers are worth writing next to what the same step costs you now in hours or invoices. The comparison is usually the whole argument.

Assembly is the only line where automation carries no credibility risk. That is why it should be first, and often the only thing you automate.

That is the honest claim and it is smaller than what most of this category promises. Automated assembly does not improve any single ad.

What it changes is the count. Published benchmarks put the winner share at 5 to 8 percent, per Motion's analysis of 550,000+ Meta ads. That is roughly fourteen tested creatives per winner.

A team producing six a month is not failing at creativity. It is under-sampling, and under-sampling looks exactly like a creative problem from the inside.

So the question to ask about any automation is not whether the output is impressive. Ask whether your monthly count of tested creatives goes up.

If the count does not move, the tool changed nothing that reaches results, however good the demo looked. Cutroom is built to move that count and nothing else.

The two purchases to refuse are the script generator sold as a strategy, and the pipeline that produces a finished ad with no human input at all.

Both fail the rule at the top of this page. Both fail in a way that is invisible until an audience sees it.

The disclosure question is separate and real. Platform policies on synthetic media differ and change, particularly for realistic depictions of people. Check the current rules where you run. Assume labelling.

On the transformation side the shape is fixed and worth knowing. Three minutes of source per take. 9:16 output. Six caption packs, adjustable inside each. Word level as the finest edit you can ask for.

The workflow needs one input from you. A person willing to speak to a camera. That is the format's requirement rather than the tool's, and it is the input that makes everything after it worth having.

Everything downstream of that take is transformation. The transcript comes from your audio. The cut comes from your words. The footage lands on phrases you chose. The captions read from the same text you can correct.

Nothing in that chain produces a claim you did not make. That is what makes it safe to automate, and it is why the credibility risk stays at zero while the count goes up.