Every piece of content comparing composer fees to AI music generation that I've seen does the same thing: it quotes the headline composer fee — £800, £2,000, £5,000 — and compares it to a subscription price. This is not a useful comparison. It ignores most of the actual cost.
The sticker price of the music is the opening number. What production teams actually spend on music for a campaign includes revision cycles, turnaround time built into schedules, stems fees, versioning work, and in some cases licensing tail costs that land weeks after the campaign runs. Comparing a composer fee to a platform subscription without accounting for these is like comparing the price of a flight to the cost of a holiday.
Here's what the cost structure actually looks like, from the production side.
The real composer engagement cost for advertising
Take a fairly typical scenario: a 30-second spot for a UK retail brand, brief confirmed, first cut delivered from the agency. The production company needs bespoke original music — not library, not sync-licensed, written to picture. They engage a mid-tier advertising composer based in London.
The fee structure typically looks something like this:
- Composition fee: £1,200-£2,500 for a 30-second original piece, depending on the composer's profile and the agency relationship
- Revision 1: usually included in the original fee — but this assumes a single round of consolidated client feedback, not three separate feedback loops from three separate stakeholders
- Revision 2+: most composers' contracts specify revisions beyond the first or second as billable. Additional revision fees run £200-£600 per round depending on the scope
- Stems delivery: if stems were not in the original brief and the editor needs them (which they usually do, for reasons covered in our earlier post on this), this is typically a separate charge: £150-£400
- Version cutdowns: a 15-second version and a 6-second version from the 30-second master, each requiring re-editing at the composition level, adds another £300-£700
A single campaign with a 30-second hero spot plus two cutdowns, going through three rounds of feedback with one round billable, requesting stems delivery: the total is realistically £2,500-£5,000 before any licensing considerations. For a 60-second spot with the same revision profile, the range is £1,800-£4,500 on composition alone — which is where our earlier post's figure came from. These are industry-realistic ranges, not outliers.
The invisible cost: calendar time
Composer fees are only part of the equation. The other part is turnaround time and what it costs when it slips.
A good mid-tier advertising composer has a four-to-seven day turnaround for a first draft. This is reasonable — composition takes time, and rushing it produces worse output. But when that first draft is delivered and requires revisions, the revision turnaround is typically 24-72 hours per round. Three rounds of revision on a two-week production schedule eats the better part of a week of calendar time.
If the music delivery slips relative to the picture lock, other post-production processes pause or proceed without the final audio. Sound design that was supposed to be mixed against the music gets mixed against a temporary track. The mix session has to be rebuilt when the music arrives. In broadcast production, this is a real cost — a senior audio engineer's time in a mix suite in London runs £150-£400 per hour depending on the facility, and rebuilding a mix session against revised music is a half-day job minimum.
These schedule costs don't appear in any "composer vs AI" comparison because they're absorbed into the general production budget overhead. But they're real, and they accrue on every campaign where the music revision cycle touches the broader post-production timeline.
Where library music sits in this picture
Production music libraries offer an alternative that seems much cheaper: £300-£900 for a sync + master license on a track that already exists. No revision cycle, no calendar dependency on a composer's schedule.
The real cost of library music is different from the sticker fee, but for different reasons. The hidden costs here include:
- The time spent searching — a music supervisor or producer spending 3-4 hours finding the right track from a library catalogue, across multiple briefs, represents meaningful overhead
- Broadcast upgrade fees if the campaign goes to television (covered in our sync licensing post)
- Stems fees if the library offers them, often matching or exceeding the base license cost
- Content ID friction for online distribution (covered separately)
- The creative compromise of using a track that wasn't written for your picture — it fits approximately, not precisely
Library music is often the right choice. We're not arguing against it as a category. But the "cheap and easy" characterisation doesn't hold up across the full cost model.
What generation changes — and what it doesn't
AI music generation removes or reduces some of these costs and does nothing about others. Being honest about which is which is important.
Where generation materially changes the economics:
- The revision cycle changes from 24-72 hours per round to minutes. If a brief produces something that needs adjustment, regeneration with modified parameters is not a scheduling event — it doesn't hit the production timeline.
- Stems are native outputs, not an add-on charge. Every generation includes the layer set.
- Version cutdowns (15-second from 30-second, etc.) are generation tasks, not composition tasks. Generating a 15-second version with the same brief parameters and a length constraint is a different order of effort than asking a composer to re-edit their work.
- There's no Content ID fingerprint in a generated piece, which removes the whitelist/dispute cycle for online distribution.
Where generation doesn't change the economics:
- Brief quality. A bad brief produces bad music regardless of whether it's going to a human or a model. The front-end creative work is not automated.
- Audio engineering. The generated stems still need to be conformed, mixed, and mastered for delivery. This is the same work regardless of where the stems came from.
- Music supervision judgment. Deciding whether generated music is right for a particular brief, whether it serves the creative intent of the picture, whether the client will respond well to it — none of this is automated. That editorial function stays with the producer.
Where composers remain the right answer
We're not saying AI generation replaces composers. There are specific contexts where a human composer is the right choice, and being clear about that matters more to us than overselling what generation does.
A brand film with a specific melodic theme that needs to carry across multiple pieces — a recognisable motif — requires compositional intent that text-brief generation doesn't reliably produce. The iterative, conversational creative process of working with a composer, where the brief evolves as the music takes shape, produces something qualitatively different from parameterised generation. For prestige campaigns where the music is central to the creative concept, not just supporting it, a composer brings craft and artistic judgment that a generative system doesn't replicate.
The comparison that's actually worth making is not "composer fee vs subscription cost." It's "which production contexts benefit from the composer's specific contribution, versus which contexts are primarily constrained by revision turnaround and stems availability?" For the latter — which is a large proportion of advertising and content music work — the economics of generation are genuinely different from what the sticker-price comparisons suggest.
