Most downloader pages only explain what they can do. This one lists what this tool refuses to do, what it returns when it refuses, and which of those answers are worth acting on. Every number below was read out of the code that is serving this page, not estimated.
Why a page about limits exists at all
A tool that only advertises success pushes every failure back onto you. You paste a link, something goes wrong, and the page either spins forever or says the link is invalid — when the link was fine. You then spend ten minutes re-copying a link that was never the problem.
So every failure here is classified into one of nineteen codes, and each code answers three questions: whose problem it is, whether retrying can help, and what HTTP status it comes back with. That last part matters if you are a script or an assistant rather than a person, because a bare 429 and a bare 502 call for opposite reactions.
The one-sentence version: 12 lookups and 60 downloads per address every 10 minutes, 200 MB and 30 minutes per file, a 15-minute download link, and nineteen named failures — never a bare "failed".
The fourteen limits
Each line gives the value, then the constant it comes from. If a value here ever disagrees with what the tool does, the constant is right and this page is wrong — please say so.
1Link lookups per address — 12 every 10 minutesEach time you paste a link and ask us to read it, that counts as one lookup. The window is a fixed ten minutes, not a rolling one, so a full window resets all at once. config.ts → RATE_LIMIT_RESOLVE_MAX = 12, RATE_LIMIT_RESOLVE_WINDOW_MS = 600000
2File downloads per address — 60 every 10 minutesSeparate budget from lookups. Reading a link and downloading the file are counted apart, so a download never eats a lookup. config.ts → RATE_LIMIT_DOWNLOAD_MAX = 60, RATE_LIMIT_DOWNLOAD_WINDOW_MS = 600000
3When you hit a limit — HTTP 429 with a retryAfterSeconds fieldThe response body carries the exact number of seconds to wait, so you do not have to guess. Retrying before that number is up costs another attempt and changes nothing. errors.ts → AppError(code, detail, retryAfterSeconds); errorCodes.ts → RATE_LIMITED: 429
4Largest file we will hand over — 200 MBMeasured on the actual file, not on what the post claims. A post advertising a 4K cut that comes out at 240 MB is refused with TOO_LARGE rather than truncated halfway. config.ts → MAX_FILE_BYTES = 209715200, confirmed in .env.production
5Longest video we will hand over — 30 minutesFacebook hosts very long streams — the reference probe hit a 4,524-second esports broadcast. A single transfer that long would hold one of only two transfer slots for over an hour, so long renditions are refused up front with TOO_LONG instead of being accepted and killed mid-download. config.ts → MAX_DURATION_SECONDS = 1800, confirmed in .env.production
6Lookups running at the same time — 2A third visitor waits up to 2.5 seconds for a free slot. Past that the request fails with SERVICE_BUSY, which is a queue that ran out — not a bad link. config.ts → MAX_CONCURRENT_RESOLVES = 2; resolveService.ts → SLOT_WAIT_MS = 2500
7Downloads running at the same time — 2The relay is the expensive part, so it gets its own smaller budget. This is why a busy moment feels slow rather than broken. config.ts → MAX_CONCURRENT_TRANSFERS = 2
8How long a lookup result stays fresh — 5 minutesPaste the same link again within five minutes and you get the stored answer instead of a new lookup — which also means it does not cost you a lookup from your twelve. config.ts → RESOLVE_CACHE_TTL_MS = 300000
9How long a download link works — 15 minutes, and only in the same sessionThe link is signed and tied to your session. After 15 minutes it returns EXPIRED. Opened in a different browser or after clearing cookies it returns NOT_FOUND, because we can no longer tell it was yours. config.ts → DOWNLOAD_TOKEN_TTL_MS = 900000; tokens.ts → HMAC payload with exp
10How long a resolved job stays in memory — 30 minutesAfter that the job is gone and the download button stops working, even if the 15-minute link had time left. The shorter of the two always wins. jobs.ts → new TtlCache<Job>(30 * 60_000, 2000)
11How many renditions we measure per post — 3A Facebook post can expose a dozen formats. We probe the first three properly and label the rest as unmeasured rather than quoting a size we did not read. config.ts → SNIFF_MAX_VARIANTS = 3
12How much we read to measure a file — 1.6 MB per lookup, within 12 secondsFacebook gives us the codec and resolution but usually not the file size. A bounded ranged read lets us read the true total out of the Content-Range header. The budget is what stops measuring from becoming a full download. config.ts → SNIFF_MAX_BYTES = 1600000, SNIFF_BUDGET_MS = 12000
13How long we wait on Facebook — 30 seconds, retried twiceFacebook is often slow rather than down. Two retries with the same 30-second ceiling; past that the answer is UPSTREAM_TIMEOUT and the problem is on their side, not yours. config.ts → RESOLVE_TIMEOUT_MS = 30000, RESOLVE_MAX_RETRIES = 2
14What we accept as a link — Facebook video, Reel, watch and share links onlyA profile, a photo, a Story, a comment permalink or any other site returns UNSUPPORTED_LINK. A fb.watch or share link whose redirect chain has expired returns SHORTLINK_DEAD instead, because the link itself is gone rather than wrong. errorCodes.ts → INVALID_URL / UNSUPPORTED_LINK / SHORTLINK_DEAD
The nineteen error codes
This is the part that makes the page worth reading. A failure that only says "failed" forces you to guess, and the usual guess — try again — is wrong for most of these. Retrying an unsupported link burns your allowance and gets the same answer.
The three groups are not decorative. Yours means changing the link can fix it. Facebook's means the link was fine and the source said no. Ours means the failure happened on this machine.
1INVALID_URL · HTTP 400 · Yours · retry: NoThe text you pasted is not a web address at all, or it is missing the part that identifies a post. Editing it will not help; copy the link again from Facebook's own share menu.
2UNSUPPORTED_LINK · HTTP 422 · Yours · retry: NoIt is a real Facebook address, but not one that holds a video we can reach — a profile, a page, a photo, a Story or a comment permalink. The address is valid and the answer will not change.
3SHORTLINK_DEAD · HTTP 410 · Yours · retry: NoA fb.watch or share link whose redirect chain no longer resolves. The reference probe saw HTTP 400 and an infinite redirect loop from stale codes. Ask for the original post link instead.
4LOGIN_REQUIRED · HTTP 403 · Facebook's · retry: NoThe video exists but is restricted to logged-in or group members. The link is fine. We do not log in on your behalf and never will — there is no version of this that we could do honestly.
5CANNOT_PARSE · HTTP 502 · Facebook's · retry: Yes, onceFacebook served a page shape our extractor does not recognise. Nothing is wrong with the link. This is the one source-side failure where a second attempt genuinely helps, because the page may render differently next time.
6NO_VIDEO · HTTP 422 · Facebook's · retry: NoThe post is reachable but holds no video we can find. A text post, an album of photos, or a link preview all land here.
7NO_SINGLE_FILE · HTTP 422 · Facebook's · retry: NoThere is video, but only as separate video-only and audio-only streams, so there is no single file to hand over. This is deliberately not reported as NO_VIDEO: a 2560p stream was measured on a real Reel, and saying there was no video would have been false.
8SOURCE_REJECTED · HTTP 502 · Facebook's · retry: RarelyFacebook's CDN refused every rendition we offered. Usually regional or temporary. If it persists on the same post for hours, the post is effectively unavailable to us.
9NOT_FOUND · HTTP 404 · Facebook's · retry: NoTwo different situations share this code: the post has been deleted or made private, or a download link was opened in a different session than the one that created it.
10UNAVAILABLE · HTTP 503 · Facebook's · retry: LaterFacebook is answering, but telling us the content is not currently servable. Waiting is the only useful response.
11TOO_LARGE · HTTP 413 · Ours · retry: NoThe file is over 200 MB. We measure the real size before offering it, so this is a final answer for that rendition, not a guess.
12TOO_LONG · HTTP 413 · Ours · retry: NoThe video runs over 30 minutes. Deliberate: relaying a longer one would hold a transfer slot for most of an hour on a two-core host.
13RATE_LIMITED · HTTP 429 · Ours · retry: Yes, after the stated waitYou have used your allowance for this window. The response says exactly how many seconds remain. Retrying earlier just spends attempts.
14SERVICE_BUSY · HTTP 503 · Ours · retry: YesBoth lookup slots were taken and none freed up within 2.5 seconds. This is queueing, not a fault, and it usually clears in seconds.
15UPSTREAM_TIMEOUT · HTTP 504 · Ours · retry: YesWe waited 30 seconds, twice, and Facebook did not answer in time. The link is probably fine; the source was slow.
16EXTRACTOR_FAILED · HTTP 502 · Ours · retry: OnceOur extractor exited badly in a way we could not classify. Retrying is reasonable; reporting it as a bad link is not, which is why it has its own code.
17INTERNAL · HTTP 500 · Ours · retry: OnceSomething broke on our side that is not the extractor. Worth one retry, then it is ours to fix.
18EXPIRED · HTTP 410 · Ours · retry: No — resolve againThe 15-minute download link ran out. Paste the original post link again to get a fresh one.
19CANCELLED · HTTP 499 · Ours · retry: YesThe lookup was abandoned before it finished — by the page being closed or the request being withdrawn. Nothing failed.
If you only remember one line: the codes worth retrying are CANNOT_PARSE, SERVICE_BUSY, UPSTREAM_TIMEOUT, EXTRACTOR_FAILED, INTERNAL and CANCELLED, plus RATE_LIMITED after the stated wait. Every other code gives the same answer the second time.
Three things this tool will not do
1We do not log in to FacebookLOGIN_REQUIRED is a real answer, not a bug we plan to fix. Private and members-only videos stay out of reach, and there is no honest way around it.
2We do not strip watermarks that were not thereIf Facebook served the file with a watermark, that is the file. We hand over what the source gave us and do not claim otherwise on the page.
3We do not guess a file size we did not measureOnly the first three renditions per post are measured. The rest are marked unmeasured rather than given an estimate that looks like a fact.
What is kept and what is not
1The video fileNever stored. It is relayed through as you download it. That is also why a download link expires in 15 minutes — there is nothing sitting on disk to serve later.
2Your link and the resultHeld in memory for 30 minutes so the download button works and so pasting the same link again is free. Then it is gone.
3Nothing about your identityNo account, no email, no login. We see a client address for rate limiting and nothing else that identifies you.
Questions people actually ask
What are the actual limits of this tool
Fourteen of them, and they are all listed above with the value and the source constant: 12 link lookups and 60 downloads per address every 10 minutes, a 200 MB file ceiling, a 30-minute duration ceiling, 2 concurrent lookups and 2 concurrent downloads, a 5-minute result cache, a 15-minute download link, a 30-minute job lifetime, 3 measured renditions per post, a 1.6 MB and 12-second measuring budget, and a 30-second upstream timeout retried twice.
Which failures are worth retrying
CANNOT_PARSE, SERVICE_BUSY, UPSTREAM_TIMEOUT, EXTRACTOR_FAILED, INTERNAL and CANCELLED are worth another attempt. RATE_LIMITED is worth retrying only after the number of seconds in the response. Everything else is permanent for that request: retrying INVALID_URL, UNSUPPORTED_LINK, SHORTLINK_DEAD, LOGIN_REQUIRED, NO_VIDEO, NO_SINGLE_FILE, TOO_LARGE, TOO_LONG, EXPIRED or NOT_FOUND changes nothing and only spends your allowance.
Why is there no error message that just says failed
Because an unclassified failure was the actual problem we set out to avoid. An unknown error reported as your link being wrong is worse than no message at all: you go and re-copy a link that was perfectly fine, and the real cause never gets looked at. Every failure path here is classified into one of nineteen codes, and each one names whose problem it is.
What does each HTTP status mean here
400 means the text was not a link. 403 means Facebook requires a login. 404 means the post is gone or the download link belongs to another session. 410 means the link expired or the short link is dead. 413 means the file is too large or too long. 422 means the link is understood but holds no usable video. 429 means slow down. 500 and 502 are our fault. 503 is us out of capacity or Facebook unable to serve. 504 is Facebook too slow. 499 means you cancelled.
Can I download a private Facebook video
No, and LOGIN_REQUIRED is the honest answer rather than an error to work around. A video that requires a login to watch requires a login to fetch, and we do not accept credentials.
Why does a download link stop working after 15 minutes
The link is a signed token with an expiry stamped into it, tied to the session that created it. After 15 minutes it returns EXPIRED. Opened from a different browser or after clearing cookies it returns NOT_FOUND instead, because the signature no longer matches a session we know. Pasting the original post link again issues a new one.
Why is there a 30-minute limit on video length
Because the transfer is relayed live, not stored. A 75-minute stream would occupy one of two transfer slots for over an hour on a machine with two cores and no swap. Refusing up front with TOO_LONG is more honest than accepting the job and killing it at the halfway point.
Why are only three renditions measured
A post can expose a dozen formats. Measuring each one means a ranged read against Facebook's CDN, so there is a hard budget of 3 renditions, 1.6 MB and 12 seconds per lookup. Anything beyond that is shown as unmeasured rather than given a made-up size.
Does the same link cost a second lookup
No, if you paste it again within five minutes. The result is cached for 300 seconds and served from memory without touching your twelve-lookup allowance.
What happens when both lookup slots are busy
A third request waits up to 2.5 seconds for a slot to free. If none does, it returns SERVICE_BUSY with HTTP 503. That is queueing rather than a failure, and it normally clears within seconds.
Do you keep the videos or my links
Videos are never stored — they are relayed through as you download. Your link and the resolved result sit in memory for 30 minutes so the download button works, then they are gone. There is no account and no email address.
Paste a link and see the real answer
If it fails, you will get one of the nineteen codes above — not a spinner.
Open the downloader