Pika repriced without changing a price

Pika's pricing page looks more generous than it did in June. Every plan costs what it cost. Every paid plan includes more credits than it did. The top-up prices that used to live in the app are printed on the page, and the expiry rules that used to be unstated are stated. By the standards the AI Credit Index uses to score disclosure, Pika got better this quarter. And a 720p clip costs 28% more than it did.

Both things are true, and the second one is the point of this essay. Since the first edition went up, Pika’s entry has been among the most-visited product pages on the Index itself, so it is worth taking the September relaunch apart properly: what changed on Pika’s own pricing page, what that does to the price of a video, why a company would make exactly this set of moves, and what it does to revenue and margin when you run the two shapes through our own AI Pricing Calculator.

What changed on the page#

The Index holds two archived reads of pika.art/pricing: 7 June 2026 and 21 September 2026. The relaunch landed between 13 and 21 September. Side by side:

	June 2026	September 2026
Free	80 video credits a month	0 credits, “packs only”
$10 a month	Standard, 700 video credits	Starter, 900 credits
$35 a month	Pro, 2,300 video credits	Creator, 3,150 credits
$95 a month	Fancy, 6,000 video credits	Fancy, 8,550 credits, slider to $880
Annual billing	20% off	20% off
Top-up price	In-app only	60 credits per dollar on $25 to $500 packs, 66 on the $1,000 pack
Plan credits expire?	Unstated	Yes, each billing period
Purchased credits expire?	Unstated	Never

Three things happened at once. The plans were renamed at unchanged prices. The allowances rose by 29% at the bottom, 37% in the middle and 43% at the top. And the unit was renamed from “video credits” to “credits”, which is the move that makes everything else legible only if you also read the burn table.

In June, Pika’s own table priced a five-second 720p clip on its Pika 2.5 model at 20 credits. In September the table lists third-party models: a five-second 720p clip costs 33 credits on Wan 3.0 and 122 credits on Seedance 2.5. Printing the table is to Pika’s credit, because it makes the arithmetic possible: Starter buys 27 Wan clips a month or 7 Seedance clips; Creator 95 or 25; Fancy 259 or 70.

The arithmetic#

Put the two tables together and the “more credits” headline inverts.

Per $10 Starter plan	June	September	Change
Credits per dollar	70	90	+29%
Price per credit	1.43¢	1.11¢	−22%
Credits per 720p five-second clip, cheapest listed model	20	33	+65%
Clips per month on the plan	35	27	−23%
Price per clip	28.6¢	36.7¢	+28%

That is the whole trick, and it is the same trick I described in The price didn’t change. Your bill did. with one refinement: Pika cut the price per credit at the same time as it raised the credits per clip. A buyer who checks the credit price sees a 22% cut. A buyer who checks the plan sees 29% more. Only a buyer who counts videos sees the increase, and videos are the thing being bought.

Two honest caveats. The models are not the same, so this is a comparison of the cheapest clip on offer each time, not of identical output; Wan 3.0 in September may well be a better clip than Pika 2.5 was in June. And Pika’s June top-up price was never public, so the top-up comparison below uses September’s rate for both.

The top-up rate itself deserves a line. Plan credits now cost 1.11 cents on every tier, the same rate from $10 to $880, which removes the odd June shape where the bigger plans were slightly worse value per credit (1.43, 1.52, 1.58 cents going up the ladder). Top-up credits cost 1.67 cents on every pack from $25 to $500, a 50% premium over the plan rate. The $1,000 pack’s 10% bonus brings the cheapest rate to 1.52 cents, which is the +36% the Index records for Pika, about twice the whole-panel median premium of 18%. All of it is printed on the page. Pika is not hiding the refill price. It is pricing the refill as a product.

Why a company does this#

Read the moves as a demand problem and they stop looking like a set of unrelated edits.

The free tier went to zero because it was a product, not a funnel. Eighty credits a month, on a tier limited to watermarked 480p, was the price of four 720p clips on a paid plan. A hobbyist could stay there indefinitely. At zero credits, the first video costs $25, the smallest pack. That trades top-of-funnel volume for a conversion decision on day one. Whether that is a good trade depends on numbers Pika has and I do not, but the direction is unambiguous: Pika decided that free generation was costing it more than the users it was warming up were worth.

The unit was redenominated because the cost base changed. In June the burn table was Pika’s own model, at Pika’s own cost. In September the table is Seedance and Wan, which Pika buys from someone else. When your cost of goods becomes a vendor’s price list, your credit has to be able to carry a price list, and a “video credit” that meant roughly one twentieth of a clip cannot. This is the COGS argument from AI pricing is just COGS finally showing up: the meter got more granular because the cost got more external.

The allowances rose because the burn rose more. If a clip costs 65% more credits and the allowance rises 29%, the plan exhausts sooner, and exhaustion is where the top-up revenue lives. Raising the allowance makes the page read as generous while keeping the exhaustion point ahead of where it used to be.

The ceiling moved from $95 to $880. The old Fancy plan topped out at $95. The new one is a slider: 90 credits per dollar, up to $880 a month, with early access to new models from $440. A studio that needs 70,000 credits a month no longer has to be an enterprise conversation; it is a subscription. For a consumer product that is a large change in who the pricing page is for.

What it does to revenue#

The cleanest way to see the revenue effect is to hold a user’s output constant and let the shape change. Take a Starter subscriber making a fixed number of cheapest-model 720p clips a month, and price the overflow at the September top-up rate in both cases, since June’s was unpublished.

Clips a month	June bill	September bill
20	$10.00	$10.00
35	$10.00	$15.01
60	$18.35	$28.37

At 20 clips nothing changes: both allowances cover it. At 35 clips the June plan fits exactly, and the September plan is 255 credits short, which the calculator prices as three 100-credit blocks at the pack rate, $1.67 each. (Pika’s smallest pack is $25, but pack credits never expire, so the pack rate is the long-run cost.) At 60 clips both users are buying packs, and the September user is buying more of them. Same videos, same sticker, a bill between zero and 55% higher.

That is a revenue story before it is a margin story. Pika did not raise what it charges for a plan. It raised what a given amount of work costs, and put the difference in the part of the bill that is sold a pack at a time.

What the calculator says about margin#

To see margin you need a cost, and Pika’s costs are not public. So the following is a structure test, not Pika’s P&L: the two pricing shapes run through our AI Pricing Calculator with identical assumptions on everything that is not on the pricing page. You can load the September shape yourself and change any of them.

The assumptions: 1,000 paid users, all on Starter, and 5,000 free users who each use 20 credits a month in both shapes (which understates September’s margin, since its free tier is zero); 15% of paid users are heavy and make three times the typical user’s clips; 30% pay annually at the 20% discount; $2 a month of infrastructure and support per paid user; top-ups at 1.67 cents a credit in both shapes; and the same cost per clip in both, because the point is to isolate the pricing change. Usage is counted in clips on the cheapest listed model, 20 credits each in June and 33 in September.

Clips a typical user makes a month	Assumed cost a clip	June: gross margin, revenue a paid user	September: gross margin, revenue a paid user
20	$0.08	58%, $10.65	64%, $12.16
20	$0.15	38%, $10.65	48%, $12.16
35	$0.08	53%, $12.91	71%, $20.17
35	$0.15	26%, $12.91	54%, $20.17
60	$0.08	64%, $23.76	78%, $37.79
60	$0.15	39%, $23.76	63%, $37.79

Three readings.

First, at every usage level and every cost assumption, the September shape earns more per paid user and more margin. That is what moving the exhaustion point does: it converts the heaviest users’ consumption into top-up revenue at the premium rate, and heavy users are where the cost sits.

Second, the gain is largest exactly where the June shape was weakest. At 35 clips a month the typical June Starter user used the whole allowance for $10, and at 15 cents a clip the cohort runs at a 26% margin. On the September shape the same cohort runs at 54%, because the last eight clips of a typical month are bought at 1.67 cents a credit rather than included.

Third, the caveat that matters. The table assumes the same cost per clip before and after. If Seedance and Wan cost Pika more per clip than its own 2.5 model did, the margin gain shrinks and could vanish; the revenue gain does not, because the revenue gain comes from the customer’s side of the meter. This is why I would not read the relaunch as a margin grab. It reads as a company whose cost of goods became external and variable, rebuilding its pricing so that the customer, not Pika, absorbs the variance above the allowance.

What to do with this#

If you buy video generation, count clips, not credits. Pika now publishes the burn table on its own page, so clips per plan can be worked out, which is more than most of the panel allows. Write that number into your renewal file and compare it next quarter at the same spend. The 29% more credits is true and irrelevant.

If you sell it, Pika’s relaunch is close to a template for redenominating a unit without hiding it: publish the burn table, publish the pack price, state the expiry, and make the clips-per-plan arithmetic possible from the page. The Index will still log the per-clip increase, because that is what it is for. But a buyer who reads the page can work out their bill, which is the standard How to buy AI without overpaying asks for and most credit pricing fails.

If you track AI pricing, this is the case for the Index’s like-for-like rule. On the Edition 2 page, the products observed in both editions show a median bundled rate that fell from 1.6 to 1.2 cents a credit, like for like. Pika is one of the products pulling that median down, and its clips got dearer. A falling credit price is not a falling price. The Pika page carries both archived reads and the dated event, so the figures here can be checked against the pages they came from.