License fees are the smallest number in any video pipeline’s cost. The real total is an architecture story: where rendering happens, what scales with success, and what you build versus what you’re given. No vendor numbers here — they change, and the models matter more. Cost structures, side by side.
The Four Cost Centers
Every approach to programmatic video pays the same four bills, in different proportions:
- Licensing — the right to use the tool
- Rendering compute — the machines that make frames
- Development — engineering time to reach feature-complete
- Operations — the ongoing cost of keeping it all running
Optimizing any one of them in isolation is how teams end up surprised. The interesting question is always the shape of each curve as volume grows.
Remotion’s Cost Shape
Structurally (whatever the current price list says):
- Licensing: free below a small-company threshold; seat-based for creators; metered per render for automated products, with a monthly minimum — a license bill that grows linearly with your success
- Rendering compute: headless-browser frame capture means every frame of every video is server CPU — the compute curve tracks total frames rendered, forever
- Development: low for React teams doing template work; high the moment the roadmap says “editor UI,” “captions,” or “background removal” — none ship in the framework
- Operations: serverless-friendly (elastic), which converts capacity planning into billing surprises instead
The compounding pattern to notice: success is metered twice — once by the license, once by the compute — while the editor and AI features you’ll eventually need are unbudgeted build projects.
The API Services’ Cost Shape
Shotstack, Creatomate, JSON2Video, Editframe — the managed-rendering family:
- Licensing: none, typically — you pay per use
- Rendering compute: bundled into per-render/per-second pricing; simple to predict, linear to your volume, and priced by the vendor’s margin
- Development: low for automation-shaped needs; a cliff for anything interactive (no editor to embed)
- Operations: near zero — that’s the genuine value
The trade is legible: you rent simplicity at a rate that scales with your success. For sporadic or moderate volumes, it’s often the right trade. At scale, teams notice the slope.
FFmpeg / MoviePy’ Cost Shape
- Licensing: free
- Rendering compute: yours, at raw cost
- Development: the big one — timeline UI, caption styling, audio mixing, effects, export pipelines: each a project, each maintained forever
- Operations: all yours, including the codec and platform quirks
The honest framing: open-source toolkits convert money into time — your engineers’ time, at a permanently higher run-rate than most teams estimate. “Free” is accurate; “cheap” is not.
Meishe Web SDK’s Cost Shape
The platform-model alternative:
- Licensing: platform license — a fixed commercial relationship, no per-render meter moving with your volume
- Rendering compute: split by design — interactive rendering happens client-side in the user’s browser (WASM on their GPU: previews, edits, quick exports cost your servers nothing), while bulk jobs run through server-side batch rendering on infrastructure you size, from one server to a worker cluster
- Development: the editor UI, AI stack (captions, background removal, beauty), effects library, and JSON pipeline ship in the SDK — the build list above is a configuration list here
- Operations: your pipeline, standard deployment patterns (documented from single-server to Kubernetes)
The structural property: the cost curve flattens where the others steepen. User-driven rendering offloaded to clients; server spend reserved for the jobs that genuinely need it; features arriving as checkboxes instead of headcount.
The Shapes, Compared
| Cost center | Meishe Web SDK | Remotion | API services | Open-source |
|---|---|---|---|---|
| Licensing curve | Platform license | Per-render meter | Bundled per-render | Free |
| Compute curve | Client-side + bulk server | Server frames, linear | Vendor-priced, linear | Raw cost, yours |
| Dev to feature-complete | Editor + AI included | Framework + builds | Low (automation only) | Everything built |
| Ops burden | Standard deployment | Elastic (billed) | None | Full |
| What scales with success | Mostly nothing | License + compute | Per-render price | Headcount |
How to Actually Decide
Model your 12-month volume — renders per month, minutes of output, interactive users — and price each shape against it, honestly including dev time at loaded cost. Two patterns consistently emerge:
- Low or sporadic volume: the API services’ simplicity is worth its slope; Remotion’s free tier may cover a small React team entirely
- Volume or interactive product: the flat-curve architectures win within months, not years — client-side rendering and platform licensing exist precisely for this regime
The generic rule underneath both: beware any cost that scales linearly with your success while offering no scale discount. Architecture is the discount. Choose the pipeline whose bill you’d be happy to receive at 10× your current volume — because that’s the invoice you’re choosing.