L'article ci-dessous est publié en anglais. Chaque chiffre qu'il contient est le chiffre publié par l'éditeur lui-même, avec la date à laquelle nous l'avons lu.
Seven tools, one question: what does it cost to find out whether this fits your work?
Why this axis and not quality
The obvious axis for a ranking like this is output quality, and we are not going to pretend to rank it. We do not generate footage and score it, and anyone who tells you they have run a fair comparison across a field this wide on the same prompts is describing a much larger piece of work than a listicle.
The axis we can rank honestly is the one every buyer hits first and almost nobody publishes clearly: how much can you make before you pay. It separates this field into five bands, from allowances that refill every month down to a tier that Google's own page marks Not available.
We considered ranking on whether each tool publishes a public API and dropped it, because we have not read and dated an API page for all seven and would have been ranking on something we could not show you. An axis we cannot source is an axis we cannot defend.
Why two avatar tools are in a video ranking
Two of these seven, Synthesia and HeyGen, sit under AI Avatars rather than AI Video in our taxonomy, because a presenter reading a script is a different purchase from a model generating a shot. They are here because this page asks a buying question rather than a capability one, and a team deciding how to get a video made is choosing across both.
What that costs the ranking is precision, and it is worth saying so plainly. The units do not convert. Ten minutes of finished video, sixty minutes of media, three clips and a one time grant of 125 credits are four different things, and no vendor here publishes a rate between them. So read the result as bands rather than seven steps: ranks one to three grant an allowance that refills every month, rank four is a single grant that does not refill, rank five states a trial grant without a quantity, rank six prints zero credits with paid packs only, and rank seven prints Not available. Inside the first band the order is our judgment about how much of the product you can evaluate for nothing, and each entry says what that judgment rests on. A stated grant with no number sits above a free tier the vendor prints as zero, because what you can spend is more than nothing even when the vendor will not say how much. The Pika review is where the zero free grant is read against the $10 entry price and the terms that sit behind the credits.
What is missing from this list
Kling AI belongs in this category and is not here. Every URL we tried returned an HTTP 446 from a bot wall with no page body, on two different hosts. That is not the same as a tool with no published pricing, and putting it at the bottom of a ranking on that basis would have been a claim we could not support. It is absent rather than ranked.
OpenAI Sora is absent for a different reason: it is gone. OpenAI discontinued the Sora apps on April 26, 2026 and shut down the Sora API on September 24, 2026. This page asks what it costs to find out whether a tool fits your work, and nobody can find that out about a product that has been withdrawn. Ranking it low would have scored an absence as though it were a measurement. Its profile carries both dates and the document they came from.
Where a vendor's pricing page will not tell us what a month-to-month subscription costs, we leave the entry price off the tool's profile entirely rather than publish the yearly rate as though it were monthly. That used to apply to Pika, whose page loads with yearly already selected, and to HeyGen, whose page served us a different language both times we asked. It no longer applies to either: Pika now prints the monthly rate in the same table as the annual one, and HeyGen's page names the currency and the cadence in its own data. It also used to apply to Runway and to Descript, and no longer does for the same reason: both cards print the monthly figure beside the annual one, and each profile says how that was settled. The Runway review is where the card itself is read against the credits and the training license.
How to read the ranking
Position one is not "the best AI video generator". It is the tool that lets you find out the most, for nothing, before deciding. For a team with a budget already approved, that ordering may be close to irrelevant, and the profiles linked from each entry carry the figures you would actually decide on.