AI Video Extender Free: What a Free Extender Actually Hands Back
Nine free AI video extenders rank for this phrase. Run here on 20 October 2026 with a marked 3-second clip, not one of them produced an extended clip that could be downloaded and measured: four demanded an account, three generated nothing, one sat in a queue of 28, and one was not a tool at all. Not one page says whether the file you get is the new seconds alone or the original plus the new seconds.
- Model
- AI video extenders, free tiers
- Observed
Nine free AI video extenders rank for this phrase. On 20 October 2026 a marked 3-second clip was put through every one that had a tool, and not one produced an extended clip that could be downloaded and measured. Four stopped at an account — three behind a login dialog, one behind a sign-up form — and on none of those four did the clip leave the machine. One took the clip, uploaded it, and then did nothing at all for six minutes after Generate was pressed. One put the job in a queue of 28 people. One never got as far as a running job because an unclosable crop dialog covered the button. One of the nine is not a tool. And the question that matters before you trust any of them — is the file you get the new seconds alone, or the original stitched to them — not one page answers, in any of the sentences it publishes.
Measured on 20 October 2026 on one Windows machine, in a browser with its own byte-counting proxy attached, using the pages Google's US results return for this phrase. Every quotation below is the page's own sentence; everything else is what happened when the clip was handed to the page.
The question this page is about
An extender and a generator are not the same promise. A generator makes a clip from nothing. An extender takes a clip you already have — a shot that is too short for the edit, an eight-second generation that needs to land a beat — and adds motion to it. That makes one question unavoidable before you spend a free allowance on it: when it finishes, what is in the file?
There are only two answers that would help an editor, and they are different files:
| What you get back | What that means in an edit |
|---|---|
| The new seconds on their own | They drop in after your clip as a separate cut, and you control the transition |
| The original clip and the new seconds as one file | You get a longer clip, but you lose the join point |
Neither is wrong. But a free tier is usually the tier where you find out, and a page that never says which one it produces leaves you to discover it after the wait.
What the pages say about the file they hand back
Nine pages were read in full on 20 October 2026. These are the only sentences on any of them that touch the output:
| Page | Its own words about the output |
|---|---|
| magichour.ai (rank 1) | "Fast, simple workflow — Upload once, extend by 4 seconds, download." / "Use the extended segment as a natural ending or transition." |
| senzia.cc (rank 2) | "Preview the result, download it in full quality, then run the extender again on your output to extend the video even further." |
| vidpexai.com (rank 3) | "Chain extensions to keep building length, preview watermarked on the free tier, then export watermark-free HD on paid plans." |
| viyou.ai (rank 4) | Nothing about the output file. Its model line says "Up to 30s". |
| easemate.ai (rank 7) | "Now, you can preview the new frames and download them with no watermark." |
| jogg.ai (rank 8) | "Choose a 5- or 10-second extension, then click Extend Video to generate the continuation." |
| vidle.org (rank 9) | "Extend in short increments and review continuity at each stage before final export." |
"Download the new frames" and "use the extended segment" are the closest any of them comes, and neither settles it: a frame is not a file, and how to use a segment does not say what the segment contains. The chaining sentence on vidpexai and senzia implies you can feed the result back in, which says something about the loop but nothing about the granularity of what comes out.
That absence is the finding. It is not that the pages hide the answer — it is that none of them states it, so the first thing a free tier gives you is not seconds or a credit, but an experiment you did not ask for.
What happened when the clip was actually handed over
The clip is 3.000 seconds long and 136,931 bytes, and it carries a marker inside its own bytes so that if it is uploaded, the bytes can be found leaving. Each page was opened in a fresh browser profile with the counting proxy in front of it, the clip was placed in the page's own upload box, and the page was then left to do whatever it does.
| Page | State after the clip went in | Outcome |
|---|---|---|
| magichour.ai | Started immediately, and opened a queue: "Queued", "28 people ahead of you", "Estimated wait: 21m 45s", with a prompt to "Sign up free to save it and get extra credits" | No output inside the window; queue never drained |
| senzia.cc | Opened a "Welcome back" dialog asking for email, password or a Google sign-in, over the tool | Stopped at the account screen |
| vidpexai.com | Same login dialog, same point in the flow | Stopped at the account screen |
| viyou.ai | Opened an unclosable "Crop Video" dialog, 2.50 s total, over the Create button | Never reached Generate |
| easemate.ai | Opened a login-and-offer dialog on Generate; the proxy recorded no clip-sized upload at all | Stopped at the account screen |
| jogg.ai | Showed the clip in its upload box, uploaded it (455,892 bytes upstream to its object store), then nothing for 369 seconds after Generate | No progress, no error, no output |
| vidle.org | Opened a "Create an account" form asking for name, email and password | Stopped at the account screen |
| kling.ai explore page | Not a tool: a roundup of prompts and samples that ends by teaching how to make images | Nothing to run |
| freebeat.ai | Served a "Vercel Security Checkpoint" instead of its content | Not measurable |
The two rows where the clip demonstrably went out are worth separating from the rest. On jogg the upload is a measurement, not an inference — the proxy weighed 455,892 bytes going to the page's object store, and the clip appeared in the page's own preview. On easemate the opposite is a measurement too: the largest single host received 16,723 bytes upstream across the whole session, which is page traffic, so a file that is 136,931 bytes did not go anywhere. Everywhere else the dialog appeared before anything could leave.
The queue is the part a free tier actually meters
Magic Hour is the only one of the nine that started work without asking for an account, and what it gave in exchange was a position in line. The dialog reported 28 people ahead and an estimated wait of 21 minutes and 45 seconds for a single 3-second clip at what the page calls its free 4-second extension. Its page also states, in the same block, "You have 3 generations left" — so the free allowance is not unlimited work, it is three jobs, each of which waits behind everyone else's.
That is a different kind of cost from a watermark or a credit. A watermark lets you judge the output and decide whether to pay to remove it. A queue makes you decide before you have seen anything. And the page's own instruction for the free path is "Sign up free to save it" — the save step sits behind an account as well.
What this does and does not establish
It establishes, as of one day on one machine, that the free tiers of these extenders mostly do not let an unregistered visitor produce a file at all, and that the one that did start work put it in a long queue. It does not establish that no free extender can output a single extension-only clip — several here may, for a signed-in account with credits, and that is precisely the configuration this run did not enter.
It also does not establish what any of them outputs. The absence is uniform enough to be worth stating plainly: across nine pages, seven of which publish several hundred words each about extending, not one writes down whether the delivered file is the original plus the addition or the addition alone. A reader who needs that answer should treat it as something to test on the first free job rather than something to read off a marketing page.
Where the pages were unreadable — one serving a security checkpoint, one replaced by a roundup that does not run a tool — the row says so instead of being filled in by inference.
The site's own ledger of AI video free tiers — each figure carrying the channel it was seen on, the date it was observed and whether it came from a vendor statement, a report or a single observation — is published as CSV, JSON and Markdown on the data page.
Where the record of this run is kept
The clip, the byte ledger for each session, the page text each quotation was taken from and the screenshots at each step are dated and stored together, so the states in the table above can be traced to the run that produced them. Nothing on this page is an average, an estimate or a projection, and no cell is filled in from a neighbouring row.