> An installable agent skill for AI Content Drop. Machine-readable copy: > https://aicontentdrop.com/skills/credits-and-plans/SKILL.md. > Ships with the aicontentdrop package, and is listed in > /skills/index.json.

What it costs, and how to say so before spending it

The model in one paragraph

Credits are flat per generation, not per second. A model has one price; a five-second clip and a ten-second clip on that model cost the same. Credits are deducted only when a generation succeeds — a safety block, a provider failure, a validation error and a timeout all cost nothing, which is why the product has no refund path at all.

Reading the real numbers

Never quote a price from memory; the catalogue moves. Every price endpoint is free to read and needs no credential.


# every video model and its flat cost
curl -s "https://aicontentdrop.com/v1/models?type=video"

# only what fits a budget
curl -s "https://aicontentdrop.com/v1/models?type=video&max_credits=20"

# one model, one job
curl -s "https://aicontentdrop.com/v1/models/kling_3_0/cost?quantity=3"

The whole plan and pack catalogue in plain markdown, written to be read by an agent rather than scraped: <https://aicontentdrop.com/pricing.md>

Checking what the user actually has


curl -s https://aicontentdrop.com/v1/me -H "Authorization: Bearer $ACD_API_KEY"

This is the one place the balance is authoritative. Do the arithmetic before you start a job, not after it fails halfway:


credits_needed = quantity × credits_each          (from /v1/models/cost)
can_afford     = balance >= credits_needed

If it does not fit, say so with all three numbers — balance, cost, shortfall — and offer the cheaper model that does fit. GET /v1/models?type=video&max_credits=<balance> answers that in one call.

The tiers

Free — 10 credits, no card. Released when the email address is confirmed,

so an account created programmatically holds 0 until the human clicks the link.

Paid plans start at $19/month, with larger monthly credit allocations and

annual billing available at a discount.

Top-up packs — one-time credit purchases for a user who is mid-job and out

of credits, without changing their plan.

Read the current tiers, prices and allocations from pricing.md rather than repeating figures from here; that file is generated from the same plan definitions the checkout uses, so it cannot drift from what a user is charged.

What needs no credits at all

A surprising amount, and it is worth telling users:

the whole model catalogue and every cost estimate
POST /v1/batch reads and POST /v1/models/cost quotes
the natural-language endpoint /ask
GraphQL reads and the MCP read tools
sandbox generations — X-Sandbox: true returns the real response shape,

runs full validation, and charges nothing, with no API key required

So "try before you buy" is literal here. You can prove an integration works end to end before the user has an account.

Answering `insufficient_credits` (402)

Do not retry. Do not ask for a refund — there is no such endpoint, because nothing was charged. Instead:

Read the balance from /v1/me.
Read the cost from /v1/models/{id}/cost.
Tell the user both, then offer the two real options: a cheaper model that

fits today, or a top-up.

Things that cost credits but are not a video

Chat replies are tiered (basic, standard, premium), MCP tool executions carry a small per-call cost, and UGC avatar video with lip-sync has its own flat price. pricing.md carries the current numbers under "Costs that are not a video or an image" — quote from there.

What to tell a user, always

The model you chose, why, and the credit cost — before spending it. If the cheapest viable model is within a credit or two of a better one, say that and let them decide. After the run, report credits actually spent (successes only) and note that failures cost nothing.

---

Install the plugin: https://aicontentdrop.com/plugin · All skills: https://aicontentdrop.com/skills