2026-10-04 · VideoFreeTier

Runway Free: 125 One-Time Credits, Converted Into Seconds

Runway's free plan is a one-time deposit of 125 credits that does not expire, and the same page prices Gen-4.5 at 12 credits per second of generated video — so the free tier converts into seconds on the vendor's own numbers: about ten seconds, or two five-second clips with five credits left over. Runway free credits and Runway ML free access are the same deposit; there is no daily refill to wait for and no expiry date to beat. That makes Runway the only model in this dataset whose free tier can be priced without measuring anything. It is also the first whose free plan says which models it will run: four of them carry a FREE label in the model picker, which means they spend none of the 125 credits at all. No Runway generation has been measured here: the 12-credits-per-second figure is the vendor's stated rate and the four FREE labels are a reading of the picker — neither is a deduction this site observed, and that grade difference is stated rather than smoothed over.

Nothing here requires a sign-up. Every dated figure this site holds is published as CSV, JSON and Markdown on the data page — no account, no email gate, nothing to register for before you can read a number.

The pricing-page rows, in the vendor's own wording

Four of the five Runway rows in the dataset come from the official pricing page, read on 2026-10-02:

Two of those carry the page's phrasing verbatim, because the wording is doing work the numbers cannot. The plan card reads *"Free forever... 125 one-time credits to explore Runway's AI tools"* — "one-time" is on the card, not inferred from it. The FAQ underneath says *"the Free plan includes a one-time deposit of 125 credits that doesn't expire"* — so the two facts that decide what this free tier is (it does not refill, and it does not expire) are both stated by the vendor, in the same sentence.

The cost row is the fourth kind of statement: the FAQ explains that credits per generation depend on model, duration and resolution, and gives Gen-4.5 as the worked example at 12 credits per second. That is a rate with three qualifiers attached, not a universal price, and the row keeps them attached.

Ten seconds, and where the tenth second goes

Put the grant and the rate next to each other and the free plan turns into a duration:

• 125 credits ÷ 12 credits per second ≈ 10.4 seconds of Gen-4.5. Nothing is estimated here; both inputs are the vendor's own published figures, and the division is ours.
• A five-second clip costs 60 credits. Two of them cost 120, which leaves 5 credits — a fifth of what a third clip needs. So the free plan buys two five-second clips and strands 5 credits.
• A ten-second clip costs 120 credits, leaving the same 5 behind. One ten-second clip, or two five-second ones: the same ten seconds either way, and the leftover is too small to spend on video at that rate.

That is the whole of the credit side in one line: ten seconds of Gen-4.5, split however you like, plus a remainder that buys nothing. If you are deciding whether the free plan is a trial or a tool, ten seconds answers it — a single ten-second generation is the entire budget, and there is no second attempt.

The expiry row changes what that ten seconds means. A grant that expires is a deadline; a grant that does not is a stock. Runway's is a stock, so the pressure on it is not time but quantity — you can leave the 125 credits sitting for months and they will still be 125 credits, and you will still only get about ten seconds out of them.

The four models that cost nothing

The fifth Runway row is a different kind of observation, and it answers the question the pricing page leaves open. Read from the model picker on 2026-10-04, four entries carry a FREE label:

So the free plan has two kinds of free in it. One is the 125-credit stock, which converts into about ten seconds of Gen-4.5 and then stops. The other is a set of models that do not draw on that stock at all — the arithmetic above does not govern them, because there is nothing to spend. Gen-4.5, the model behind the 12-credits-per-second rate, is not one of the four.

Two things about that row are worth keeping straight. First, the grade: this is a report row, read by the site owner from the picker on 2026-10-04 with a screenshot on file. No generation was run against any of the four, so no deduction was measured. The row records a label, not a measured cost of zero, and those are not the same claim. Second, the date: a label in a picker is a UI state, and UI states change without announcement. That is why the row carries a date at all, and why the next reading of the same picker is a new row rather than an edit to this one.

There is also a structural point the picker makes on its own. Seedance is ByteDance's model and Kling is Kuaishou's; neither is Runway's. A free plan can therefore route you to another company's model at no credit cost — the same lesson this dataset keeps arriving at from the other direction, that the name on the model is not the name on the bill.

Only one grant in this dataset has no clock on it

Widen the frame and Runway's deposit looks unusual among the models this site has recorded, because the others hand out something that either expires or has to be re-earned:

Read the columns, not the amounts. Three of these five grants have to be claimed again; two do not. Two state an expiry and three do not — and this table does not read a silence as "never", because a row that does not mention an expiry is a row that does not mention an expiry.

The practical difference shows up the moment you have a finished idea. With a re-earned daily grant, the free tier is a rate: you can plan against it because tomorrow brings more. With a one-time deposit, the free tier is a total: what you see is all there will ever be, and it is spent once. Runway is in the second group, and unlike Kling's expiring package it carries no deadline to force the decision.

What the rows do not say

The rows are thin on purpose, and the gaps are worth naming rather than filling:

• How long the four FREE labels will last. The picker named four models on 2026-10-04. The pricing page still describes the free tier only as *"a selection of generative creative models to test concepts"* — a phrase that resolves to no list and carries no date. A picker label is the weakest kind of evidence this dataset keeps: a UI state rather than a statement, and it is recorded as one.
• Resolution, and whether free output carries a watermark. No Runway row covers either. Both are recorded for other models in this dataset, so the absence here is a hole in the record, not a finding about Runway.
• What a generation actually deducts. No generation has been run against Runway, so there is no measured deduction: no login-wall check, no duration ceiling, nothing to set beside the vendor's 12 credits per second. The one report row Runway has is the model picker, and a label is not a deduction. For PixVerse and Wan the same page pattern ends with a measured deduction — 50 credits for a five-second generation, and 6 / 15 / 30 credits for two, five and ten seconds — and those rows exist precisely because a vendor's rate and a measured deduction are not the same kind of claim.

That last gap is the honest limit of this page. A reader who wants to know what Runway's free plan really deducts per clip should treat the ten seconds above as the vendor's arithmetic and expect the measured number to be its own row, with its own date, once someone runs one.

The record now: 74 rows across nine models

Runway's first four rows have been in the dataset since 2026-10-02 and the fifth landed on 2026-10-04; this page is the first to read them together. The dataset stands at 74 rows across nine models — Seedance 2.5 (22), Wan 2.5 (17), PixVerse (8), Kling AI (6), Veo 3.1 (6), Grok Imagine (5), Runway (5), Sora 2 (3), Hailuo AI (2). Thirty-five of those are vendor statements, thirty-eight are reports, and one is a single uncorroborated account; twelve of the reports are observations taken inside a logged-in account, and Runway's model picker is now one of them.

Scope of what this page says

The first four Runway rows were read from runwayml.com's official pricing page on 2026-10-02, from the plan cards and the FAQ beneath them; the fifth was read from the site's model picker on 2026-10-04, in a logged-in session. They describe those screens on those dates, not Runway's product in general, and the file names on this site still carry the seedance-2-5-free-tiers stem for link stability — the dataset itself covers every model named above. Comparisons on this page are made against dated rows from the same dataset: Kling AI's trial package and 720P rate (klingai.com, 2026-10-03), PixVerse's free-plan figures and its measured 50-credit deduction (app.pixverse.ai, 2026-10-02), Wan 2.5's plan-table rows and measured 6 / 15 / 30-credit deductions (create.wan.video, 2026-10-03), and Doubao's five generations a day (2026-08-11).

The counts on this page — 74 rows across nine models, five Runway rows, twelve measured rows — are counts of this site's own records, not claims about how free tiers behave in general.

Not confirmed on this page: no Runway generation has been run by this site, so the 12 credits per second is the vendor's stated rate and not a measured deduction, and the four FREE labels are a reading of the model picker rather than a measured cost of zero; the free plan's resolution and its watermark behaviour are not stated in any row we hold; how long those four labels stay is unknown, because the pricing page resolves the free tier to no list; and the 125-credit deposit, its non-expiry and the Gen-4.5 rate are read from a public pricing page on 2026-10-02 and have not been independently re-verified by a second party.

The full dataset — 74 rows, each carrying channel, observation date and source type — is on the data page in three formats, free to use under CC BY 4.0 with attribution. To turn a grant and a usable rate into an attempt count, the attempts-needed calculator uses only figures from that dataset and estimates nothing on your behalf. Related pages: PixVerse AI Free for a free plan whose measured deduction outweighs its daily grant, and Kling AI Free for a trial package that expires.

*Update cadence: figures on this page change only when the dataset gains a dated row that touches them; every figure keeps its own observation date — check it before quoting as current.*