Notice
New Update
Ongoing
2024
03-31 : V3 is delaying
We are experiencing the delay in the training of the new AI v3.
We need more time before users can enjoy the new AI.
This is hard time for us but we cannot release the new AI prematurely.
We will strive to release the new AI soon
2023
12-01 : Spring is coming
[Intro]
In this article, we are going to introduce you the incoming new AI v3 release.
The contents include :
- the executive summary
- the ETA of AI v3
- the achievement of AI v3
- the update schedule of AI v3
[Summary]
We would like to release the new version of our AI v3 in a month or so.
The new AI has massively improved.
We will update in a several phase.
[ETA]
We would like to release in December 2023 hopefully.
The exact date is unknown and we would not announce until the update will be done.
Check our website regularly.
[Achievement]
The overal quality of nudification has massively improved.
We have resolved most of quality issues observed in current AIs (v1 or v2).
In addition to that, we have achieved the personalized nudification, meaning that not only breast, areola and nipple expression seem to be realistic but the new AI can reveal each girl specific size, shape, color and texture. We has been obsessed to achieve this since the beginning of our project. AI is generally bad at generating or recognizing rare features (e.g. extremely small or large areola, puffy nipple)
Of course, we also will add the new capabilities, which many users yearned for
[Update Schedule]
We will upgrade the AI version first, which make users feel the improvement immediately.
We will gradually unlock additional features of the new AI, which include several image manipulation features, higher resolution and more.
2022
12-31 : The ETA of AI v3
[Summary]
In this article, we would like to share the ETA of AI v3 and the backgrounds.
The contents include :
- the ETA of AI v3 (short)
- the progress (short)
- the explanation of difficulties (long)
[ETA]
We hope to release the new AI (v3) somewhere in the 2023.
We are grateful to users who support our project and anticipate our new model.
A lot of users have contacted us since the last update, giving us the inquiry and feedbacks.
That gave us a lot of encouragement to tackle this difficult project.
We were the earliest believer of nudification future and have been doing the research and showing its potential.
We will 100% deliver what we promised and a substantial improvement with the new AI.
[Progress]
We have done the following things so far :
- built and augmented proprietary datasets
- trained auxiliary AI models
- experimenting AI architectures and training schemes
We are going to keep experimenting and start a production training of the AI after satisfying with experiments.
Roughly speaking, we have completed 70% of the task.
[Nudification difficulty]
Why AI v3 delayed ? Simply because it is an inherently difficult task.
Nudification is a true AI task in a sense that the task is not solvable with a computer program and nearly impossible to do even by human hands.
Nudification is not merely stitching body parts picked from arbitrary persons to depict nude, like ordinary deepfakes doing. We do not want to see such creatures.
Nudification aims for revealing hidden inner truth about a specific person.
That is exactly what we are trying to achieve and the holy grail of we mens.
[Practical Aspect]
AI development is all about data and computation, which are the source of difficulties.
The difficulties we are facing are as follows :
- difficulty of AI development in general (validation necessity about data, parameters, and programs)
- difficulty of creative AI (creative AI is more difficult than other AIs)
- difficulty of nudification AI (nudification AI is the most difficult among creative AIs)
[vs Software 1.0]
Ordinary software development validates only programs, which are easy to test or execute.
In AI development, you need to validate data and parameters too in addition to programs of a program, which costs you much more time and money. Only one test in AI development costs you hours, days, even weeks of time, and of course money in accordance with time spent.
Because of so much computation needed to train AI, you need a dedicated hardware for AI training that is pricy. For examples, you need to pay a 30 USD per hour to rent the latest GPUs from datacenters, and you need to rent hundreds of thousands of hours to create a really awesome AI.
Programs would give you apparent errors when there are bugs but AI is not. That forces you to spend additional time.
What is worse, bugs of AI are notoriously difficult to spot because AI is adaptive.
Because of this difficulty, even a lot of machine learning or AI researcher so called keep failing to reproduce their results, which they are proud of in their litarature. This is a dark side in the AI research community that outsiders never know.
If you ignore reproducibility, you cannot get any progresses like any other science or engineering.
[vs other AIs]
The difficulty of creative AI compared to other AIs comes from the boost of the factors mentioned above.
So, creative AI needs much more computation, very sensitive to parameter changes and susceptible to noises in data.
That partially explains the reason why practical applications of creative AI are relatively few.
[vs creative AIs]
The difficulty of nudification AI comes from hierarchical and complex nature of human body
It is easy for human to recogize patterns of a nipple but AI is not.
It is obvious for human that a nipple is contained in a breast, a breast contained in a upper torso and a upper torso is contained in a human.
Still, AI is bad at recognizing this and sometimes depicts a nipple in backgrounds not a human body like the outdated nudification AI did.
For a long time, the AI research community mostly focused on human face generation, typically celeb faces. That is actually far easier task. They competed with each others on that task.
Such results were hardly applicable to practical applications.
Actually, we have witnessed people who failed by just mimicking such results, even got contacts from former developers of nudification AI.
The researchers have been paying attention to human body generation recently and finally noticed its difficulty.
But, our task is nudification, which means you need to accurately reveal hidden inner truth not just depicting a human body.
Due to ethical or moral reasons, nudification or deepfake technology are prohibited in the research community, which means that little knowledge is available and you need to research by your own hands.
[Closing]
The difficulties mentioned above would explain the reason why other nudification AIs have not emerged until now.
AI development need a principled approach and hence need a rigorous and diligent working practice.
Lazy people or copycats who choose an apparent shortcut never reach to a goal in the machine learning field, because it is a trap.
We are trying to tackle to this difficult problem just by huge data and huge computation. That are the only empirically proven way to the success.
Is is a simple but the hardest to do.
We will keep fulfilling our promise as we always did.
03-07 : Hope Plan Arrived
[Summary]
Now users are able to skip poor nudified images indefinitely.
This is the much anticipated feature since you can greatly reduce the risk of quota loss for poor results (stable performance per price).
We recommend it especially for beginners.
The feature is available only with [HOPE] plan.
For more details, please visit pricing page (special price as of this writing).
02-22 : Milestones
[Summary]
In this article, we are going to introduce you the planned updates.
The contents include :
- the progress and ETA abount AI v3.0
- other updates in 2022
- reviewing user feedbacks
- futures of DeepSukebe
[AI v3.0]
The ETA of AI v3.0 would be latter half of this year.
The overview of improvements is the same as described in the other section (ONGOING section, please click top right tab in this page).
We are still in the research phase, followed by the production training phase, both of which take a lot of time and money, which is the difficulty of releasing creative AIs applicable to real world.
If you want to know the improvements in more details, please read this article further as we review user feedbacks.
In short, we aim for solving all of known problems.
[Attribute manipulation]
This new feature will be delay too.
The release would be the same time with AI v3.0 as opposed to the initial plan.
We will try releasing the feature earlier than v3.0 although we give top priority to AI v3.0.
[Redemption improvement]
We will improve the existing redemption feature to mitigate quota loss further.
Our hope is that users enjoy nudification without any stresses.
We really received a lot of feedbacks about the redemption feature or nudification failures, which we think must solve in coming years.
Our strategy is multi folded - the combination of an automatic system detection by scoring AI, quality check by human/users and more.
The planned improvement allows users to reject bad images indefinitely, which will be released as soon as possible.
[AI v2.0]
We will keep improving AI v2.0 in the meantime of finishing the research of AI v 3.0.
[User feedbacks]
DeepSukebe will evolve based on our vision (what the future world should be like) and user feedbacks, which tell us the right way to go for achieving men's dream.
We have collected a lot of feedbacks from users across the globe since 2020.
To prioritize user feedbacks, we have done a survey to get the understanding of user demands from 2021-04 to 2021-08.
That made us realize the necessity of attribute manipulation feature and accurate breast size prediction especially busty one.
The top 5 result is as follows.
- attribute manipulation option (17.5%)
- accurate breast size (17.5%)
- erase hair on clothing/body (9.9%)
- full clothing improvement (8.2%)
- pubic area improvement (7.3%)
Please refer to the picture below for more details (sorry for English only).
[User feedbacks and AI v3.0]
With AI v3.0, we will give improvements to users, which are summarized as :
- greatly reduce poor/bad nudification
- improve the quality of success nudification further
In order to do so, we categorized user feedbacks and located bottlenecks.
We will tackle every single problems or cases given by users, some of which happens in specific conditions.
The examples are :
- DeepSukebe dunno really big boobs
- body reconstruction is bad when arms or legs are crossed
- hands/finger disappearance
- I like fat bitch
- clear image
- option for pubic hair
- pubic area is blurry
We appreciate users a lot for giving us as many feedback as we cannot introduce here.
Some of users sent us sample pictures about failure case via twitter or email periodically.
Also, some had the same vision with us, requesting advanced feature, which we will tackle for next 5 to 10 years.
[Future of DeepSukebe]
The future is of course, to achieve men's dream.
In more precisely, that will be achieved by increasing the fundamental value of DeepSukebe or solving problems such as :
- meet sexual curiosity (fundamental human nature)
- nudify who never undress in front of you no matter how you are rich or have power
- with less than 0.1 USD (cheap enough to be available for everyone)
- in a safe and legal manner
We have been pondering a long term plan to achieve that because sexual curiosity of men is diverse, varying degree and evolving and a lot of obstacles/interventions are expected.
So the actualization is step by step and our project will span more than decades (please be patient).
The reason why such long time is needed comes from inherent difficulties of creative AIs and risky projects
- AI development is 3 times or more difficult than ordinary software development
- especially creative AI is very sensitive to every single parameters
- need a great deal of computational resources for awesome result (need a lot of money, increasing efficiency in hardware and software and awaiting for innovation)
- must deal with or mitigate devastating risks such as making one's life end literally
- human resources are very precious (never hail trojan horses)
- prevent critical information leakage by 3rd parties (procurement, financing, development, distribution)
These things are low profile and hard to imagine to people other than practitioners.
Still, you can grasp the risk we are facing by the failure of the creater of [Silk Road] - the former biggest crypto dark market.
The creator ended up in life imprisonment and all fortune was taken out.
That is why we need careful considerations on plans and designs and improvements is slow because agile methodology or the practice of fail early fail often do not work completely, one cereless mistake eventually leading to death
2021
11-03 : New Payment System
The payment system has come back online.
We will gradually add the rest of pricing plans although only some of plans are available due to the need of the system warmup right now.
So much thank you for patience.
[What is new ?]
- UI renewed
- Available cryptocurrency changed (BTC & XMR added, BCH dropped)
- Improved protections for payment accidents
[What happened ?]
Coinbase (our backend vendor for the payment system) suspended our account suddenly.
They changed their TOS in August possibly due to the media pressure (following the start of UK & US media campaign).
[Why it took so long ?]
We need to carefully rebuild the system to reduce payment accidents and stabilize services.
Cryptocurrency payments is error-prone due to the need of a lot of manual inputs by users.
Also, it it hard to find stable and trustworthy vendors for the payment system since most vendors do not allow services like us.
[Background]
The world is becoming more and more harsh for the porn industry and deepfake technology.
Recently, Mastercard effectively banned porn contents.
Even crypto payment processors cannot be the last resort as we have experienced.
All major SNSs like Twitter, Reddit, Medium, etc banned our accounts and are censoring users posts.
They have every reason to behave in a politically correct way.
10-25 : Payment Stopping
The payment system have been stopping since last week.
We need a few more days to rebuild the payment system.
We will make the system more stable and error tolerant for payment accidents
So much thank you for a lot of inquiry.
09-06 : Abuse Protection & Compliance
We will introduce abuse protection for nudifier.
All photos that users upload will be processed with masking/degradation except premium users.
This is to comply with international standard rules and recent criticism from media and activists.
We can no longer provide stable services without the restrictions on nudifer.
In August, our servers was down many times because they put pressure on our vendors and our accounts were suspended.
Media and activists are trying to ban and outlaw nudifier technology.
They demands us to monitor and prevent all of the abuses or misuses that users can do potentially.
That is totally ridiculous and arrogant demand as we are just researchers or tech enthusiasts.
But they have the power in today's world, unfortunately.
Our website is on the verge of shutdown.
We think their tyranny will be end some time but not near future.
Their unreasonable power is inherited from the flaw of financial system, that the tech will solve in future.
It is shame that we cannot provide full capability to all users.
08-15 : New Partner Joined
Our new partner - DeepNude.gg - opened cool website.
They use DeepSukebe AI engine and make true nudification experience more accessible across the globe.
This partnership gives users convenient payment methods and help us focusing on AI research for men's dream.
For more details, please click bottom (i) icon.
08-11 : Privacy Improvement
We are working on the task to provide better privacy and safer place.
That includes :
- Accepting privacy coin (Monero) for payment (Completed)
- Adding delete all button for premium users (Completed)
- Ensuring image privacy (Completed)
- Reducing automatic deletion time of images (Completed)
- Allowing VPN usage for all users including free mode (Completed)
The features are to meet the requests about privacy that we have received from users so far.
Also, we need to make clear the position of our app, namely we just provide contemporary digital canvas and brush to facilitate artworks by users, same with Photoshop.
Based on the above decision, we have deleted public mode for assets, also known as item page.
Next, we have set automatic deletion time to 1 day.
Also, we have updated the description about privacy in FAQ and privacy policy.
We will keep improving privacy for all users.
We believe privacy is a human right.
Please let us know if you have more requests although we may not be able to reply.
08-10 : Servers back to normal from down
All servers have been back to normal from down.
The Server was down for 6 hours from 2021-08-09 19:10:00 UTC+0.
The issue was caused by our supplier.
You can get the real time status from below.
We will increase redundancy to improve service stability.
So much thank you for a lot of inquiries.
08-07 : AI servers congested
AI servers have been very congested because of the surge of access since a few days ago.
We have release the fix to some of the problem by the congestion.
Still, it is difficult to resolve the congestion anytime soon.
We expect it gradually improves in a week or so.
07-21 : DeepSukebe v2 available for all tickets
Premium users can choose DeepSukebe v2 models now.
If you have the quota dedicated for v2 beta test, v2 beta quota will be consumed first.
Otherwise, both of v1 and v2 will consume common quota.
NOTE : Ticket for v2 beta test that has v2 beta quota was sold out.
07-19 : v2 Beta Ticket Sold Out
We will stop selling v2 beta ticket on July 19th.
Premium users can use remaining v2 beta quota after that too.
So much thank you for cooperation !
07-01 : Payment system is stopping
06-28 : Mail delivery delayed
Currently, mail delivery is delayed or even not delivered even if your email address is correct one.
It is unclear when it is back to normal because the problem is rooted in backend vendor.