Skip to main content

Filter by idea status

10000+ Ideas

Neural Filters improvementsOpen for Voting

NEURAL FILTERS are not that great.  They are certainly not what they are being touted about at present and need VAST improvement to be useful for serious professional retouching.   So my question is how long will be be before these filters really work on a profesisonal level?   My main issue is with the "SMART PORTRAIT" filter Here is an unretouched picture before filters are replied:  FIRST IMPROVEMENT REQUEST: The blue outline box is not positionable.  With many pictures, parts of the face are outside of the box, so, for instance, if skin smooting is applied, the person's necl may be outside the blue box and the filter will not apply.  The person's face may be smoothed, but the neck will be wrinkled.     The blue box should be poistionable and sizeable.   @ 2ND IMPROVEMENT REQUEST: Here is an image for which the FACIAL AGE" filter is applied with only a +5 point correction: As you can see, most of the change is to the hair, and very little to the face, and the hair style completely changes in a way that is not even age-related, and the overal all appearance makes the subject look younger, if anything, not older.   Even at +30 correction, almost no chage occurs tot he face. Facial aging only becomes slightly eveident at +40 correction, btu almost all changes occur to the hair, not the face. Here is an animated GIF.  As you can see, almost all "FACIAL AGE" change occurs to the hair, not the face!  The hair style is instantly changed with the slightest age change, and then stays the same.  THIS IS NOT REALISTIC.   HAPPINESS FILTER:  Another disappointment.  Attached below facial proprtions are way off and not realistic, and teeth look horrible:   Animated GIF: ALL of the SMART PORTRAIT Filters need a lot of improvement.  I hope it is forthcoming very soon becuase until that happens, the neural filters is, at best, a curiosity that may be fun to play with, but of very little use for serious retouching.                    

Zesty_soul0350Participant

Proposal for the problem of fake illustrations using AI: About the embedding function for working hiOpen for Voting

1. Current IssuesImage generation technology using AI has been rapidly developing in recent years, and has been used in many situations. On the other hand, there is an increasing number of cases where AI-generated illustrations are disguised as "our own-made illustrations," and this is a serious problem in illustration competitions, corporate projects, and individual requests. In particular, cases where an AI-generated image is partially processed and claims to be "hand-drawn" is difficult to spot using traditional discrimination methods.  2. Limitations of traditional measures Currently    many platforms and review agencies are trying to identify AI images, such as:      Analysis of the completed image (analysis of noise and feature patterns)      Submitting progress (but it can be disguised by intentional modification)      Human review (highly burdensome, not certain)  However, due to the advancement of AI technology, it is becoming more difficult to distinguish using simple image analysis or examination.  3. Proposed Solution To solve this problemwe will implement the function to embed work history (such as pen pressure, drawing time, layer operations) in image data" in Clip Studio Paint and Photoshop. I'll suggest it.  4. Benefits of introduction ✅ Prevention of disguise: Images without work history cannot be hand-drawn proof, making it difficult to disguise AI-generated illustrations. ✅ Creator Rights Protection: Make it easier for authorized illustrators to prove their work. ✅ Improved reliability in competitions and corporate projects: This is effective evidence in cases where hand-drawn proof is required when commercial use is required. ✅ Seamless industry standardization: When installed in major paint software, it will be easier for illustration posting sites and competition operators to make use of this information.  5. Technical implementation direction "Work history (pen pressure, drawing time, layer operation history, etc.)" is automatically saved as metadata in .clip or .psd files.      When exporting to PNG/JPEG, optionally embed work history as watermark data.      Develop authentication tools to enable submission sites and competition operators to verify whether they have a work history.  6. SummaryDue to the evolution of AI technology, the way illustration production is being changed. However, in order to protect the rights of authorized illustrators, the introduction of the "Embed Work History Function" should be a useful measure for the entire industry.  We hope that this proposal will be considered by your technical team. If you have any questions or need any additional comments, please feel free to contact us.  Sincerely,

Hoping to see a hybrid program interface, featuring cohesive design interface and ai implementationOpen for Voting

So, with the release of stargate on the horizon for next generation ai, I am hoping to save up for a 2 in 1 around that time (four years) It would be really exciting to see a hybrid program that implements firefly into a cohesive powerhouse program. Basically what I was hoping to see for this program would be a hybridization of audition, premiere and Photoshop (but with elements designed from drawn inspiration for clip studio.) It would be incredible to see the program work intuitively with multiple pages and a mix down like feature incorporated to synthesize audition and premeire.  I have a really hard time understanding premiere, but I would love to see this for projects where firefly could do things like generate solo tracks with machine learning like sound effects, tracks like audio vocals and instrumentals, and the ability to cohesively cooperate with a csp like interface. A revamped program like this would be really cool. I wanted to prose the idea that a new operating system with a quantum inspired dual scale Integer based bit relativity with 9 - -9 digits and computer language that uses math of relativity and dedicated neural processing units to emulate the probabilities as potential in the language to be more or less probable to be 1 or 0, where -1 is neutral and you could suppose things like connecting probabilities to emulate superposed digital states like 7 and -6 within the developed language Programming as an example to state their probabilities to be potentialized.   If you consider desktop oses as if like chess and mobile oses like rubix cubes, you could imagine synthesized os to be a bridge for mobile and desktop hybridiztion.  Imagine a desktop that uses this dual synthesis to have apps on desktop programs and possible new kinds of devices from the advent of ai. It would be cool to propose this idea that a game I invented called sandbox could hybridize chess and rubix cubes, where a 1x1 to 16x16x16 template could be used to rotate squares and customize rules with custom peices or not even need to use peices could be a prime example for the metaphors inspiration for the os. I'm acutely intuitive on higher dimensional geometries and inevitably I'd like to develop a smartphone in basis of this os, but you have to start somewhere!!!  +.-