Bài viết bên dưới được đăng bằng tiếng Anh. Mọi con số trong đó là con số do chính nhà cung cấp công bố, kèm ngày chúng tôi đọc được.
A free tier is a budget, not a gift, and a budget has a job. The job is not to keep you inside a product for as long as possible at no cost. It is to let you answer one question before you pay for anything: can this tool produce the thing you actually need. Of the seven products on our video ranking, six state a free tier value on a page we could read: in minutes, in credits, in a count of finished clips, or as a flat Not available. One does not: Luma Dream Machine's card did not give us a free allowance we could establish. Reading a published allowance as marketing is the mistake. Reading it as a budget with a real limit, and asking what it can and cannot buy, is the useful version of the same page.
One pass, or a loop
What decides a free tier's value is not its size, it is whether it comes back. A grant spent once buys a single look at the product. A grant that refills buys repeated attempts at the same job, which is the more useful thing to plan around.
| Tool | Free tier, as the vendor states it | One pass or a loop |
|---|---|---|
| Runway | 125 one-time credits | One pass. The grant does not refill. |
| Synthesia | Usable for up to 10 minutes of video/month | A loop. |
| Pika | 80 monthly video credits, 480p only | A loop. The 480p cap is a separate question, see below. |
| Descript | 60 minutes (1 hr) / month | A loop. |
| Google Veo | Not available | No grant at all, on the Gemini API. |
That is three distinct outcomes across five products: one pass, a loop, and no grant at all. An axis that sorted all five into the same bucket would not be worth publishing. This one does not.
Runway grants 125 credits once, not every month, and once they are spent the account keeps working but stops generating. Synthesia, Descript and Pika all refill on a calendar: Synthesia returns 10 minutes of video a month, Descript returns 60 minutes of media a month, and Pika returns 80 credits a month. Google Veo has no grant to spend on the Gemini API, which is the surface we read, and we did not check Google's consumer products for a separate way in.
Whether a grant comes back is a related question to the one our video ranking sorts on, and not the same one. The ranking asks how much you can produce before you have to pay, which is a question about size. This page asks whether the allowance returns, which is a question about repetition. A large grant that never comes back and a small one that refills every month can sit close together on the first question and far apart on this one, so read the two pages as two cuts of the same field rather than as one verdict stated twice.
What Pika's 480p cap decides, and what it does not
Whether Pika's monthly refill functions as a loop depends on what you need to finish, not on Pika. The free tier returns 80 credits a month, and every free generation is capped at 480p. If your deliverable can ship at that resolution, you get a real, repeating evaluation.
If it cannot ship at 480p, the credits still arrive on schedule and still do not move you closer to what you were asked to produce. That is not a defect in Pika's plan, it is a fit question about your own job. A tier locked to 480p can still tell you a great deal for nothing: whether the tool follows a prompt, whether motion holds together, whether the timing of a shot is close to what you asked for. It cannot tell you whether the final frame is usable in a deliverable, because the plan was never built to answer that. Ask which of those two things you need before you count the 80 credits as room to evaluate.
A published absence is still an answer
Google's Gemini API pricing page prints Not available in the free tier row for Veo, and that line is a stated fact about the offer on that surface, not a gap in what we could find. It removes the guessing: on the Gemini API there is no unpaid evaluation to plan around, so budget for a paid evaluation instead.
Had that row been left off the page, you would be guessing whether an unpublished free option exists somewhere, or whether we simply failed to find it. Not available, printed in a row built to carry exactly that answer, is worth more than an omission, because it tells you outright not to plan an unpaid pass on that surface. That is the third outcome on this page, distinct from a one-time grant and from a loop: not a smaller version of either, but the absence of both.
Whatever shape your free tier takes, spend it on the actual job, not on a tour of the product. Ten minutes of Synthesia video is enough for one real training module, not ten short experiments in different styles. Sixty minutes of Descript editing is enough for one real episode, cut the way you would actually cut it. Use the whole allowance on the version you would ship, and put that in front of whoever signs it off. A free tier that can produce your real deliverable once, at a quality someone is willing to approve, has already answered the only question it was ever built to answer.