2026-10-03 · VideoFreeTier

Wan 2.5 Free: The Plan Page Prices a Video but Not the Free Tier

Wan's official plan page can tell you exactly what a video costs and cannot tell you what the free tier gives you. Sixteen listings on that page pair a credit figure with an "Accelerate: up to N images or M videos" line — the two memberships, the six credit packs, the two membership gift cards and the six credit gift cards. Divide one number by the other and all sixteen return the same two values: 5 credits per video, 0.25 credits per image. A 30-credit pack and an 8,000-credit gift card agree to the decimal.

The Free column of that same page carries no credits-per-month figure at all. Its only credit line is "Daily check-in to earn free credits", and no amount is attached to it anywhere on the page. So the price of a video is published to the credit, and the size of your free budget is not published at all. All of it was read from create.wan.video on 2026-10-03, and every figure is a dated row in the dataset.

Then we went through the gate. Opening the generator with no account produced a login modal before a single prompt could be submitted. Logged in, the free tier gave up the number the vendor never publishes: a daily check-in earns 6 credits, one 2-second generation deducts exactly 6, five seconds costs 15, ten seconds costs 30, and twenty seconds or more is paid-only. Three deductions, one flat rate: 3 credits per second of video. So the entire daily grant buys one 2-second clip — and a 5-second clip is never fundable by a single day's grant at all.

Nothing on this site sits behind a sign-up. Every dated figure we hold 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 row. That is not true of every product in this dataset: Wan's own generator, we measured, will not run a prompt without one.

The free column, in Wan's own words

Read from the official plan page (create.wan.video/pricing) on 2026-10-03. This is the entire free column, line for line:

Four of the eight entries in that table are a documented absence, and a fifth — the credit source — names a mechanism without a quantity. What is left are three numbers: 1 concurrent video, 1 concurrent image, 6 image styles. None of them is a budget. Two are caps on how much you can do at once; the third counts styles, not output.

The free column is what the paid columns have

The Pro and Premium columns are not longer versions of the same list. They carry rows the free column does not have at all:

Read down that table and the shape of "Wan free" is not a small allowance — it is a column defined by three missing rows. The three lines that describe what you actually receive (no watermark, full resolution, longer clips) exist only from Pro upward, and so does the row that would answer the question you came with.

There is a mechanical reason this is easy to miss. A plan table is read column by column, and an absent row leaves no mark in the free column — nothing to notice, no zero to react to. The only way to see it is to read across, which is what the table above does. The dataset stores the absence explicitly, so the empty space becomes a dated, checkable row instead of a blank.

The cost side is priced to the credit

The same page is unusually precise about cost. Sixteen separate listings carry a credit figure next to an "Accelerate: up to N images or M videos" line, and dividing one by the other gives the same answer every time:

Same arithmetic on the image side returns 0.25 credits per image, at every one of those listings. One video is therefore priced at twenty images, consistently, from the cheapest pack to the most expensive.

Two cautions belong on those numbers, because the page supplies the words and not the label:

• It is our arithmetic, not a stated unit price. The page never prints "5 credits per video". It prints a credit total and a maximum output count. The rate is what falls out when you divide them.
• It is a floor, not a typical cost. The page's own word is "up to", so 5 credits is the cheapest per-video cost its numbers permit. A longer clip, a higher resolution or a different workflow will cost more — the page does not say how much more.

The dataset row carries both cautions in its note, because a derived figure and a stated figure are not the same grade of evidence even when they sit in the same table.

Which brings the asymmetry into focus. The page can price a video to the credit and cannot price the free tier at all. Cost is published; the grant is not. Anyone trying to plan around "wan 2.5 free" is holding the price list and missing the budget.

What we measured

Two things were checked directly rather than read off the page.

The gate (2026-10-03). A headed browser session opened the generator at create.wan.video/generate with no account. Before any prompt could be entered, a login modal appeared: "Welcome to Wan / Log in / Don't have an account? Sign up", with Google and GitHub sign-in options. Screenshot on file. There is no anonymous generation path. This is a report row, not a vendor row — the vendor describes a free plan; we recorded what actually happens when an unauthenticated visitor tries to use it.

The free tier's numbers (2026-10-03). Logged into the free account, these are the figures the vendor's pages do not carry anywhere:

Three deductions, one rate. Divide each by its duration — 6 ÷ 2, 15 ÷ 5, 30 ÷ 10 — and every one returns 3 credits per second of video. No bulk discount for the longer clip, no minimum charge for the short one. It is the cleanest linear rate in this dataset, and it is the number the vendor's own plan page never states. (Runway's comparable figure is vendor-stated at 12 credits per second; Wan's is measured at 3. Both rows carry their grade.)

Now put the grant next to the cost, which is what the dataset exists to do:

• The daily check-in earns 6 credits. The cheapest clip costs 6 credits. The free tier funds exactly one 2-second video per day — and only on a day you check in.
• A 5-second clip costs 15 credits: two and a half daily grants. One day's check-in never covers it. You save for three days, or you accept two seconds.
• A 10-second clip costs 30 credits: five daily grants. Since the grant arrives by check-in rather than by calendar, the longest clip the free tier allows costs five consecutive days of showing up.

That last point is the one worth carrying away. The free tier's currency is not credits, it is attendance — and the plan page's wording ("Daily check-in to earn free credits") is the only hint the vendor gives that this is how it works.

The two cost figures do not reconcile

There is a loose end, and it belongs in the open rather than smoothed over.

The plan page's sixteen "Accelerate: up to N images or M videos" lines imply 5 credits per video — 300 credits is listed as "up to 60 videos". The measured deduction for a 5-second clip is 15 credits. Both cannot describe the same generation: 5 credits per video would be a one-and-two-thirds-second clip.

The page never says what duration its "up to N videos" figure assumes, and we are not going to invent one. So the dataset holds both numbers, dated, with their grades — the vendor's ceiling arithmetic and our measured deduction — and does not average them into a value nobody published. If the Accelerate line refers to a cheaper mode, a shorter default, or something else, the page does not say, and that silence is itself the finding.

The record now: 51 rows across four models

Wan's seventeen rows land the dataset at 51 rows across four models — Seedance 2.5 (22), PixVerse (8), Runway (4), Wan 2.5 (17). Ten of the Wan rows are vendor statements read from the official plan page on 2026-10-03; seven are site-measured: the login gate, the daily grant, three deductions, the derived per-second rate and the duration ceiling. Seven of the 51 rows are a documented absence, and four of those seven are Wan's.

The practical consequence for this page is simple. Wan is now the second model in the dataset whose free-tier question is answered mostly by absence — PixVerse was the first, before its in-app plan table was read — and the first where the missing row sits beside a fully priced cost table on the same screen. It is also the clearest case so far of why the dataset carries two grades: the vendor's page supplied the price list, and only a logged-in session could supply the budget.

Scope of what this page says

The Wan rows were read from create.wan.video's plan page and generator on 2026-10-03, through interactive browser sessions: the plan table from an unauthenticated visit, the login gate from an unauthenticated attempt to generate, and the grant, the three deductions and the duration ceiling from a logged-in free account on the same day, with screenshots on file and no independent re-verification by a second party. The plan page names no model version in its tier columns; the same site's home page and generator default were promoting a later generation on the same date. So these rows describe the platform's tiers as they stood on 2026-10-03, not a per-version entitlement — which is the honest reading of a page that does not attach its numbers to a version.

Seedance 2.5 rows were read across Jimeng, Doubao, Doubao Work, Mina Labs, Lovart, BytePlus ModelArk, hosting platforms and vendor material between July and August 2026. Runway and PixVerse rows were read from their official pricing pages on 2026-10-02.

The counts on this page — 51 rows across four models, sixteen priced listings, seven absences — are counts of this site's own records, not claims about how free tiers behave in general.

The full dataset — 51 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. The file names still carry the seedance-2-5-free-tiers stem for link stability; the dataset now covers Seedance 2.5, PixVerse, Runway and Wan 2.5. To turn an allowance 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 the first model read off an official plan table, and Hailuo AI Free for why an allowance belongs to a channel rather than a model.