
Video is billed differently
Processing video: length limits, per-second billing and the result
Video is charged per second, not per file. The photo price list is no help in reading it.
Short answer
Footage is not one image but hundreds computed one at a time, and every tariff follows that arithmetic: per second, per resolution, or per credit pack. Because each frame is calculated separately, the boundary between the altered area and the untouched one shifts slightly from frame to frame — and that shift is what shows up as flicker the moment the clip plays.
Collected below is what these services write about video themselves: the durations they accept, the resolutions, the file formats and the prices. Figures they do not publish are not guessed at here. The map of what sits behind the name is on the front page.
The arithmetic of a clip
Seven seconds is hundreds of images
One number explains almost everything odd about these pricing pages: an image model does not know what a clip is. It knows frames.
Seven seconds at 24 frames per second is 168 images. Twenty seconds at 30 frames per second is 600. Each frame is a separate job for the model: computed, billed, then glued back into the sequence. From that come the two questions people ask most — why a short clip costs as much as dozens of stills, and why the result is never as steady as a still image.
Public tariffs readable before an account exists are available only from some of the services, and they are not written in the same units. The table copies what each one states, without converting them into a common measure.
| Service | What it states about video | Where it is written |
|---|---|---|
| Ainudez | 50 credits for 480p at 5 seconds, 75 for 480p at 7 seconds, 150 for 720p at 5 seconds, 300 for 720p at 7 seconds. One still image is 15 credits. | Pricing page |
| Deep Undress | MP4 or MOV under 50 MB. Billing starts from 5 seconds, and clips longer than 20 seconds are processed only for their first 20 seconds. | Upload box and in-app notice |
| Undresswith | Its smallest pack, 33,000 gems at $19.99, is described as enough for 16 videos; the largest, 800,000 gems, for 410. One still image is 100–110 gems. | Pricing page |
| WaveSpeed | Billed per execution: video at $0.01 to $0.15 per second depending on the model, images at $0.005 to $0.14. The rate moves with resolution and parameters. | Pricing page |
| Pornworks | Every request to its address is answered with a 403 and a check screen. Not one of its tariff documents can be read from outside, so there is no figure to copy here. | Direct request to the site |
The Ainudez row is both the most open and the most explanatory. Moving from 480p to 720p at the same length triples the cost, from 50 to 150 credits. Adding two seconds at 720p doubles it again, from 150 to 300, although the duration only grew by 40 per cent. And those 300 credits equal twenty still images on the service's own price list. Seven seconds of mid-resolution footage costs what twenty photographs cost.
The same pattern reads out of Undresswith in different units: 33,000 gems covers 16 videos or roughly 300 stills at the rate it publishes itself. Again one video is about twenty images. At WaveSpeed, which rents third-party models instead of running its own, the bill arrives directly as a price per second — the underlying cost structure shown naked, with no pack in the way. Where those models come from, and how much any of them explains about the engine underneath, is a separate page.
One figure appears in no table anywhere: the price of a failed attempt. Ainudez warns that every execution is counted, including when you are only fixing a face. So the real cost of a clip is the list price multiplied by the number of retries — and that is never one. How the packs themselves are read is covered in the piece on how far the free tier goes.
What shows up on screen
Five faults that only appear because it moves
A still image hides small errors; motion exposes them. None of these five will ever show in a screenshot, and all five show on playback.
Flicker between frames
Two consecutive frames are two separate jobs, so skin tone, shadow and texture are never identical. The eye is extremely sensitive to differences that small when they arrive 24 times a second.
A pulsing mask edge
The replaced region is redefined on every frame. Its boundary drifts by a few pixels back and forth, and along a shoulder or a waist that drift reads as a trembling line.
Objects crossing through
A hand, a bag strap, hair falling forward — anything passing across the processed region disappears, breaks or melts in the frames where it crosses, then reappears intact afterwards.
Thin fabric and patterns
Stripes, lace and sheer material change shape every frame because the model re-guesses what is behind them. In a still there is one guess; in a clip there are hundreds of different ones in a row.
A face that is not the same a second later
Features drift even when nobody asked them to. That is why many services sell a face fix as an extra paid step — and why it does not solve the problem, because it also works frame by frame.
All five come from the same cause: there is no memory between frames. As long as the price is charged per frame or per second, no service has an economic reason to compute the second frame while looking at the first.
From send to download
Six things that decide how a clip turns out
The sequence is nearly identical across services, and at each step there is a decision they make for you without saying so.
The queue and the waiting time
A video job occupies a graphics card far longer than a single image, so services put it in a queue. Out of that come two extra products with separate prices: acceleration and membership status. Deep Undress, for instance, states that speeding up a job costs three more balance units, while its membership is promised to save 10 to 20 per cent per execution.
What nobody states is how long that queue is without acceleration. A figure in minutes never appears in the documents, only in the advertising. As long as there is no written number, paid acceleration cannot be compared with anything: you are paying to fix a condition whose size was deliberately left unpublished.
Why the limit is 20 seconds
Deep Undress says it outright: clips longer than 20 seconds are still accepted, but only the first 20 seconds are processed. This is not an editorial policy, it is a cost ceiling. Twenty seconds at 30 frames per second is already 600 model jobs, and raising that number raises the compute bill the service itself pays to its graphics provider.
The other limits work in the same direction. File size is held under 50 MB, the format is narrowed to MP4 and MOV, and billing starts from five seconds even if your clip is shorter. None of the three is about output quality; all three are about how many frames are allowed inside one price.
The resolution promised and the one returned
The Ainudez price list stops at 720p. That matters, because nearly every phone recording now sits at 1080p or above. Sending 1080p footage to a service whose top tier is 720p means the result comes back downscaled, and that reduction is never announced as a loss — it simply goes unmentioned.
WaveSpeed takes a different approach and states that its price moves with resolution and parameters. That is structurally more honest, and the consequence is the same: high resolution is not an included feature but a price multiplier. In every case, resolution is the first thing cut when costs have to come down, because it is the one limit that does not show immediately on a phone screen. What else is left unstated about the output, down to resolution and watermarks, has a page of its own.
Where the clip goes
None of these services computes on your device. The file moves whole to their servers, is split into frames, processed, reassembled and returned through a link. That means the recording sits on somebody else's infrastructure for the whole process, and how long it stays there is decided by their documents, not by you.
The answers they write differ, and are not always consistent inside a single service. Undresswith states 24 hours encrypted and then deletion, while account, payment and IP records are stated as kept for seven years. Deep Undress gives three different figures on three different screens. The detail of that, along with what remains after the tab is closed, is in the page on what happens after login.
Footage in front of Philippine law
Philippine law does not distinguish photographs from video when it comes to sexual material produced or shared without consent. The Anti-Photo and Video Voyeurism Act of 2009 names photo and video coverage together and prohibits copying, selling, distributing, publishing or broadcasting such material — expressly including through the internet and cellular phones — with three to seven years of imprisonment and a fine of ₱100,000 to ₱500,000.
Section 12 of the Safe Spaces Act of 2019 adds the broader online provision, covering media with sexual content uploaded or shared without the victim's consent. And Section 6 of the Cybercrime Prevention Act of 2012 raises the penalty by one degree when the act is committed through information and communications technology. All three apply to the same file, and the person carrying them is the one who pressed send, not a service whose servers stand abroad.
Why a free sample proves nothing
The opening balance these services hand out is the size of one image, not one clip. Undresswith gives 100 gems for verifying an email, while its own tariff puts one still at 100–110 gems. WaveSpeed gives a one-dollar credit with no card, and at the top of its video rate that dollar is gone in under ten seconds of footage. Neither balance is enough to see a complete clip come out of their system.
So the decision to pay is always made on a sample the service chose itself: short, already compressed, and recorded under the conditions friendliest to the model. What each one actually publishes, row by row, is in the comparison of what each one publishes, and the money risks that follow — including the charges that will not cancel — are collected separately.
Common questions
About processing footage
Answered with what can be defended: frame arithmetic and the documents these services publish themselves.
Why is video so much more expensive than a photo?
Because what is counted is not the clip but every frame inside it. On the Ainudez price list, seven seconds at 720p takes 300 credits while one still takes 15 — a twentyfold gap, which is exactly the amount of work being run.
What is the real maximum length processed?
It depends on the service, and only some write it down. Deep Undress states that it cuts at 20 seconds: longer clips are still accepted, but the rest is ignored. The others announce no limit at all, and with no written figure nothing can be promised here.
Does raising the resolution remove the flicker?
No. Higher resolution makes the drifting mask edge more obvious, because detail that used to drown in compression becomes visible. Only a model that takes the previous frame into account removes flicker, and that is not an option on any menu.
Why does the same clip with the same settings come out differently?
Because every execution starts from a different random point, and in footage that difference is multiplied across hundreds of frames. The practical consequence is financial: two attempts on the same file are billed twice in full, with no discount for a repeat.
Is the clip processed on my phone?
Not by any service on this page. A model that size needs a data-centre graphics card, so the file always travels to their servers first. Any claim to the contrary is contradicted by the size of the app's own installer.
Is processing someone else's footage against the law in the Philippines?
The Anti-Photo and Video Voyeurism Act names photo and video coverage together, and the Safe Spaces Act covers media with sexual content shared without the victim's consent; the Cybercrime Prevention Act raises the penalty by one degree when ICT is used. This is an editorial page, not legal advice: for a concrete case, talk to a lawyer.
Did this desk test the video output itself?
No. We processed nobody's footage and sent files to no service. Everything here comes from the documents and pricing pages they publish themselves, plus frame arithmetic you can repeat with a calculator.