Makeroll guide

A recorded pilot. Every line itemized.

Most AI UGC cost talk is estimates. This page is anchored on one recorded pilot with a metered ledger: 47 calls, 3 concepts, 6 exports.

What the 47 calls actually bought

The pipeline, itemized

3concepts generated and stored, each with its own prompt record
6platform-adapted exports, all passing technical QA
91–96QA gate scores, 6 of 6 exports passed the release gate

Research calls decompose the brief and gather references, metered per call, logged with timestamps. A larger run in the same pipeline, the Makeroll-UGC campaign, produced 16 final videos, all 16 passing the same technical QA. The cost model scales roughly linearly with export count because the metered calls are the cost.

The recorded figure covers one pilot's metered research and generation spend across 47 calls, checked 21 Sep 2026. It excludes Codex access, human time, subscriptions, media spend, taxes, and retries.
Exclusions

What the pilot figure does not include

Anyone quoting an AI UGC price without this list is selling something.

  • Model access. Codex or equivalent coding-agent access is a subscription above the metered calls.
  • Human time. Writing the brief, picking concepts, running critique. Real cost, not metered, not zero.
  • Subscriptions. Any flat-rate tooling used alongside the metered calls.
  • Media spend. Distribution and ad delivery. Production cost and placement cost are different lines.
  • Taxes and retries. Failed calls get re-run. The ledger records the successful pilot; a real week includes friction.
Against traditional UGC rates

Directional, with sources

RouteTypical cost per videoTurnaroundBasis
Traditional UGC creatorA per-video fee; platforms and agency listings commonly quote a band for a single delivered UGC adDays to weeksPublic creator-market listings and agency rate cards, checked 24 Sep 2026
MakeRoll documented pilotMembers or partners only (47 calls, 3 concepts, 6 exports)Same dayPilot ledger, 21 Sep 2026

The honest comparison is not "free versus paid". It is throughput and repeatability: a creator delivers one video per brief on their schedule; a documented pipeline delivers concepts and exports on yours, with every cost line recorded. You still pay for taste. That is the part the club teaches.

The honest baseline

Median AI TikTok · 418 views

Riffkit's 2026 dataset of 1,504 AI-generated TikToks puts the median at 418 views. That is the real floor for AI content without distribution and without craft. We cite it because promising otherwise would be lying. The studied exceptions are documented and dated, checked 22 Sep 2026 (the hook structure behind them):

  • 1,727,992,414 combined plays across three public reels, recomputed from platform counters.
  • 35M+ views on Qatar Airways' AI film, per Google's published case study.
  • 130,000 personalized ads, 94M views for Cadbury's SRK My-Ad, per WPP's writeup.

None of these are member outcomes. They are evidence the ceiling is real when the craft is.

Why it repeats

Repeatability beats one-off cheap

One recorded pilot proves the cost model. It does not make you good at this. The repeatable part is the loop: one brief in, critique, one finished roll out, every week, with the ledger telling you exactly what each attempt cost. That is what the weekly club is for. Membership is open now (see the plans).

These are references, not promised member outcomes.

The camera is optional. Taste is not.

One brief, one finished roll, critique each week. Ledger documented. Sources checked 21–24 Sep 2026.

Keep reading