Per minute, per deliverable, or retainer?
The same month of video work quoted three ways, showing where each model earns and leaks, plus a rule for choosing on clip length and revision risk.
Quote the same month of work per finished minute, per deliverable, and as a retainer, and the three numbers will not be close. On short-form work they can differ by a factor of four, and the direction of the gap flips entirely once the deliverables get longer.
That is not a rounding issue. The unit you quote in is a bigger decision than the rate you attach to it, and most freelancers pick the unit by copying whoever they read last.
The job we are quoting
One month for a local fitness studio. Fixed scope, no ambiguity:
- 10 vertical clips, averaging 20 seconds each
- 1 brand film, 2 minutes
- One consolidated revision round on the batch
- 9:16 for everything, plus 1:1 versions of three clips
Total finished runtime: 200 seconds of short-form plus 120 seconds of brand film — 320 seconds, or about 5.3 minutes. Eleven deliverables. Fourteen exports once ratio variants are counted. Each pricing model bills against a different one of those three numbers.
Quote A: per finished minute
Per-minute quoting for AI video freelance work is commonly advertised somewhere in the $50–$150 per finished minute range. Take $100 as a working midpoint.
5.3 minutes × $100 = $533.
That is the whole month — ten clips, a two-minute brand film, a revision round and fourteen exports — for five hundred and thirty-three dollars.
The per-minute model was built for a world where finished runtime correlated with work. In a corporate explainer or a training module it roughly does — a twelve-minute video really is about four times the work of a three-minute one. Short-form breaks that correlation completely. A 15-second hook takes as much concepting, briefing, review and export handling as a 90-second piece, and sometimes more, because the shorter it is the more the first two seconds have to be right.
Per-minute quoting on short-form is not aggressive pricing. It is a category error.
Where it earns: long single deliverables with low concept density. A 12-minute training video at $100/minute is $1,200 for one concept, one script, one approval chain and one export.
Where it leaks: anything under a minute, anything with more than one concept, and anything where revisions attach to individual clips rather than to the whole runtime.
Quote B: per deliverable
Per-deliverable rates for the same work are advertised across a wide band: low hundreds for a short social cut, four figures for something with real production behind it. A modest pack discount on five or more is the usual convention.
Price the ten short clips at $200 each and the brand film at $600 as a heavier deliverable:
- 10 × $200 = $2,000
- 1 × $600 = $600
- Subtotal $2,600, less 15% pack discount = $2,210
That is four times the per-minute quote for identical work.
Per-deliverable pricing bills against concept count, which is what actually consumes your time. It also gives you a clean unit for every downstream conversation: a kill fee counts deliverables generated, a scope change adds or removes deliverables, and "just one more" is a question with a published price attached.
Where it earns: short-form, multi-concept work — most of the market. Also anywhere the client is likely to add scope, because the additions have a rate.
Where it leaks: deliverables of wildly uneven size under one rate. If one price covers both a 15-second hook and a 3-minute brand film, you are subsidising the second with the first. Band them, as above. The other leak is revision risk: a per-deliverable rate with unbounded revisions decays with every note, so cap the round in the same document as the rate, per batch and not per video.
Quote C: monthly retainer
Freelance AI video retainers are commonly advertised from around $1,500/month upward; small-business social management starts in a similar place and tiers up from there. Productized agency AI-UGC programmes sit well above both, priced in the thousands a month and scaling steeply with clip count.
Quote this job at $1,800/month, with scope capped at twelve deliverables, one revision round per batch, and one longer piece.
Against a single month that is $410 less than the per-deliverable quote. Against twelve months it is a different business: predictable revenue, one negotiation instead of twelve, and a client who stops evaluating each video as a purchase. You accept a lower per-unit rate for demand certainty and scope-creep risk for admin savings. The scope-creep half is what kills retainers priced at the bottom of the band, which is why the cap belongs in the document and not in your head.
Where it earns: predictable ongoing demand, and anywhere the relationship is worth more than the month — a retainer client tells you about their roadmap; a project client tells you about their brief.
Where it leaks: spiky demand. A client who needs 4 videos in January and 25 in February pays you too much once and too little once, and the second month is the one you remember.
The three quotes side by side
| Per finished minute | Per deliverable | Retainer | |
|---|---|---|---|
| Bills against | Runtime | Concept count | Time |
| This job | ~$533 | ~$2,210 | $1,800/mo |
| Best fit | Long-form, single concept | Short-form, multi-concept | Predictable ongoing demand |
| Worst fit | Anything under a minute | Uneven deliverable sizes under one rate | Spiky demand |
| Revision exposure | High — notes do not add runtime | Medium — capped per batch | High — the cap is the only defence |
| Scope-change handling | Renegotiate | Add a line at the published rate | Swap, do not add |
| What the client argues about | Whether the cut was long enough | Whether that counts as one video | Whether they got value this month |
The last row is worth reading twice. Every pricing model creates a specific argument, and you are choosing which argument you would rather have. Per-minute clients push for longer cuts, which is a bad incentive. Per-deliverable clients push on the definition of a deliverable, which is manageable if you defined it. Retainer clients push on monthly value, which is what reporting exists to answer.
Revision risk is the variable most freelancers underweight, and it hits the three models unequally. Under per-minute, revisions are free to the client by construction — a re-cut does not add runtime, so it does not add fee. That is the model's structural flaw and no rate solves it. Under per-deliverable, revisions are bounded if you wrote the cap and unbounded if you did not; the definition that works is one consolidated round per batch, applied in a single pass, with anything beyond it billed as a new deliverable. Under a retainer, revisions are absorbed into the fee, which is fine until the client discovers there is no marginal cost to asking. The fix there is the same cap, stated per batch.
One production detail changes the arithmetic under all three. Versely's editor renders preview passes at 480p for free — with a short per-user cooldown, so it is not an unlimited loop — and charges once for the final export regardless of clip count. Applying a consolidated notes round and re-checking it at preview resolution is therefore cheap, as long as the export happens once at the end. What it does not change is your time, which is what the cap protects.
Everything else on Versely bills in credits with no free plan or free allowance behind it, so a month of variants is a real spend to forecast before you quote. Estimating credit cost before dispatching a batch is the pre-quote check, and the worked scenarios for a finished 30-second ad and a 30-second talking-head video give you a floor to build a rate on.
The rule
Three questions, answered in order:
- Is the average deliverable under 60 seconds? If yes, do not quote per minute. Short-form runtime has almost no relationship to the work involved, and per-minute pricing on it will underquote by a multiple.
- Is the client's demand predictable month to month? If no, quote per deliverable. A retainer against spiky demand loses on the heavy months and creates resentment on the light ones. If yes, and the volume clears roughly eight to ten deliverables a month, a retainer beats per-deliverable on both admin and relationship value.
- Is revision risk high — new client, unclear brand, committee approval? If yes, quote per deliverable with an explicit round cap, whatever the other two answers were. It is the only model where the cap has teeth, because exceeding it produces an invoice rather than a conversation.
Short version: long-form and single-concept, per minute. Short-form and multi-concept, per deliverable. Predictable ongoing volume with a written scope cap, retainer. When two of them fit, quote per deliverable first and convert to a retainer after three months, when you have data on what the client's month actually looks like rather than what they said it would.
FAQ
Can I quote per minute for short-form if I raise the rate?
You can, and it still misprices the work — you have moved the error, not removed it. At a rate high enough to make ten 20-second clips profitable, a single 90-second piece becomes absurdly expensive and the client notices. No rate fixes a wrong unit.
How do I convert a per-deliverable client to a retainer without dropping revenue?
Set the retainer at roughly what their last three months averaged, cap the scope slightly below their heaviest month, and give them the swap rule — they can substitute a queued concept for a new one but not add. Converting on an average rather than a peak keeps the number honest; the cap keeps it profitable. The scope side is in the local business content retainer.
What counts as one deliverable when a client wants three aspect ratios?
Whatever your scope document says, which is why it needs to say it. The workable definition: a deliverable is a concept, and ratio variants of the same concept are included up to a stated number. Without that sentence, an eleven-deliverable month silently becomes a thirty-three-export month.
Do these advertised rate ranges hold in practice?
They are asking prices from rate cards and marketplace listings rather than audited transaction data, and nobody publishes fill rates — how many operators actually book work at these numbers. Use them to sanity-check the shape of a quote, not to justify one. How agencies price AI video services covers the tier ladder above this, and the plan and credit pricing page is the input side of whatever rate you land on.