Batch Video Processing Service — Pricing Proposal

1. Your cost basis

The engine (Very Good FFmpeg) bills per GB processed = input bytes + output bytes, exactly the model Vidrotation already uses (built/verified in the repo: estimate.js, DEFAULT_PRICE_PER_GB = 0.50, output sized at a typical 1080p CRF-23 bitrate). Engine rate cards (public, 2026):

Blended engine cost (what you actually pay as your aggregate volume grows across all users):

That is $0.007/GB/month — basically a rounding error versus engine compute. Amortized fairly, across ~20 paying users it's ≈$0.70–0.90 per user/month. It behaves as a fixed platform cost (hosting your batch outputs and source pools), not a per-unit driver. Treat it as a covered line item inside each plan, not as a metered input.

2. Critical finding: naive pass-through is unprofitable at scale

Two levers fix it — and you should use both:

• Price operations, not raw GB. Users think in "convert this", "make a thumbnail" — charge per operation, which lets you bundle the thumbnail's input-read cost into a fair flat fee.
• Mark up the credit pool and monetize overage. Sell a monthly credit bundle priced from your blended cost with healthy markup; charge more on overage. Like all SaaS video APIs, most users don't consume their full bundle (60–80% typical), which is where the margin actually lives.

3. Unit economics per operation

Design point: a typical 60-second 1080p clip ≈ 500 MB input. Engine runs input+output GB for each op. Verified engine costs at the $0.50/GB entry tier:

Markup logic: ~2× on engine cost, sized so that even a worst-case job (thumbnails) holds ~50% GM. That buffer absorbs failed-job retries, low-balance retries, and your $14/mo S3 before it ever erodes margin.

4. Pricing scenarios

Scenario A — Pure per-GB credit (naive; shown for contrast)

Only viable if you budget on 60% or lower utilization and lean on overage income. Fragile — one heavy user on Agency erases the tier.

Scenario B — Per-operation flat fees (user-friendly, + S3 as fixed platform cost)

Charge per operation at the §3 prices (convert $0.99, resize/crop/trim $0.79, compress $0.69, thumbnail $0.50, thumbnail free with any same-batch paid op). GM ≈ 50–58% per op. S3 + your platform cost are recovered by a small platform/export fee per batch cover (source pooling + output hosting):

Best for discovery and simplicity — the buyer knows the price before they click. Predictable revenue per job; no need to guess utilization. Recommended direction for the "batch interface with trim/crop/thumbnail/convert/resize/export" you described.

Scenario C — Recommended hybrid: operation credits in a monthly bundle

Marry A and B: the plan includes a pool of operation credits (1 credit = 1 standard op, thumbnails ~0.4 credit), priced from your blended engine cost, with markup and overage. Storage/S3 lives in the plan.

Because engine cost per op drops with your aggregate volume ($0.167→$0.087/GB), your blended cost per bundle falls as you grow while the sell price holds — margin widens at scale, the exact opposite of Scenario A.