What one credit costs on Luma
Luma publishes credits per second by model and resolution and never states what a credit costs. That is the closest near miss in this register: the difficult half of the conversion is done and the easy half is missing. As of 2026-09-22.
| What is published | As the pricing page carries it |
|---|---|
| The consumption rate | Credits per second, by model and by resolution |
| Price of a credit | Not stated on the page |
| What that leaves | A rate in an internal currency with no exchange rate |
| Distance to a ledger row | One published figure |
Inclusion rule. What the pricing page states about credits and what it omits. An omission is listed because it is the reason this channel has no ledger row. Order. What is published first, then what is missing.
1Where Luma sits in this column
| Channel | Price of a credit |
|---|---|
| Hedra | Not stated, but the plans divide out |
| Luma | Not stated, though the rate is per second |
| Pika | Not stated, and the rate is per feature |
| PixVerse | 225 credits to the dollar, one reference |
| Runway | One cent, stated beside the rate |
| SceneMixer | $0.0149 down to $0.0078 by pack |
Inclusion rule. Channels that meter in credits. Whether the credit price is published decides whether anything on that page converts into money, so both answers are listed. Order. Alphabetical by channel name.
2The hard half is done and the easy half is absent
Working out how many credits a second of video consumes takes a seller real effort: it differs by model, by resolution and sometimes by feature. Luma has published that. What is missing is a single number that the seller sets by fiat and already knows.
Which makes this the most frustrating shape in the column. Every other unpriced channel here is missing something structural. This one is missing a sentence.
3Why the register refuses to estimate it
Plan prices exist, so a credit price could be inferred by dividing a subscription by its allowance. The result would be a number, and it would silently assume that the whole allowance gets spent and that no credit expires unused.
Both assumptions are wrong for most buyers, and a derived figure carries none of that doubt with it into a comparison. A blank cell is less useful and cannot mislead, so the ledger keeps the blank.
4What a buyer can do with half a rate
The published half still ranks Luma's own models against each other, which is the decision a buyer already committed to this channel actually faces. It is the cross-channel comparison that fails.
For anyone still choosing a channel, the practical move is to treat the missing figure as a reason to shortlist elsewhere and come back if it appears. Rates get published; this one has a smaller distance to travel than most.
- What Luma publishes instead of a rateCredits per second, with no credit price
5Sources
Read from the pricing page published at lumalabs.ai on 2026-09-22. The rate this column sits beside is on Luma; what the register accepts as published is on the method page. The same column on other channels: Pika, Hedra. Every column: what each channel publishes.