pancakes

MicrostockGroup Sponsors


Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - mino216

Pages: [1] 2 3
1
Shutterstock.com / Re: Shutterstock "Contributor Fund"
« on: January 30, 2024, 13:30 »
But they won't need the images again. The AI places don't care, because they don't need to use the images again. Just like what they scraped from the Internet. Once the image has been used for machine learning, it is not being used again.

What I'm saying is, the six year license is for training, not use. Once the training is done... we get nothing.

They will. They will create new AI models and these models will need the images once again. How do you want to do that without these images? The new models will increase amount of parametres, will be trained to do more stuff that the old model was not trained to, you need these images to train it on. You cannot just use the previous model that will miraculously use the previously trained information in a completely new algorithm with new features. You do not need it for fine-tuning the algorithm, but in the six-year period, there will be at least 3, maybe even more totally new generations of models trained completely from scratch.

2
Shutterstock.com / Re: Shutterstock "Contributor Fund"
« on: January 08, 2024, 10:54 »

For example, the Shutterstock's deal with OpenAI is for 6 years. After that, OpenAI has to licence the images again.

This means OpenAI will have access to new submitted photos for 6 years. After 6 years they will not get any new photos. There is no need to licence old photos that were already used to train their AI again.

They will need to licence them again, because they have licenced them for 6 years only. They have to retrain the system periodically including the old images. AI algorithms do not work so easily that you can just "add" the new stuff on something you already have. DALL-E 4 will use the same images to be trained on as DALL-E 3, it needs them again, and will have some new too. And if you do not have a licence to use the old images after X years, they will have to get the new licence even for the old ones.

3
Shutterstock.com / Re: Shutterstock "Contributor Fund"
« on: January 05, 2024, 16:49 »
I am surprised people kept thinking they would get infinite and regular payments for this. Once an AI is trained, it's trained, it doesn't need your images every month or every couple of months.

But it is totally obvious. Imagine yourself as Shutterstock. Would you offer these images forever? And never get any other money again? That would be financial suicide. You would offer them for some time period. The AI algorithms is developing as well, it needs more images, newer images,... you need to retrain the new AI model,...

For example, the Shutterstock's deal with OpenAI is for 6 years. After that, OpenAI has to licence the images again.

4
Why do you think that? The income per image used for training on Adobe is much smaller than for selling a license for normale usage. In my case on Adobe is was less than a cent for an image. How am I supposed to earn more from AI training than from selling my real photo licenses in the future like this?

I have explained that in my post, have you not read it? Once again, you cannot say "The income per image used for training on Adobe is much smaller than for selling a license for normale usage" because YOU DO NOT KNOW what is the ratio between AI and normal images. If Adobe sells normal images (for example) for 1B USD and AI images for 10M USD per year, then AI images are earning you more if they are more than 1/100 of income of your normal images. Why should it be higher when the market is smaller? It does not make sense. How can Adobe pay contributors the same 1B USD for AI images per image per year if AI images have sold for 10M only during the same time period? If your normal images will sell for 1000 USD per year but AI images generated by system based on your images for 10 USD year (because the market is much smaller), why on earth should you get the same income per image per year when these images have not sold for such amount of money? I do not get it.

Also, don't you think the market for AI image generatores will be saturated pretty soon? Do you think new AI engines will keep popping out for the next 100 years? Some of the AI engines like modjourney are already creating almost perfect results in almost every topic. Their demand for new images (apart from the fact that they do not pay us for training anyways!) for them is minimal.

Therefore there should be payment for each sold image generated by such a system (which is a problem). I am not saying all these systems are OK for contributor.

They might need new images every couple of years for some things like technology where appearance changes fast. But even if I popped out a million  of new photos of new cell phones and electric cars each year and got paid for each image for training by Adobe I would still be making significant less than what I used to earn till now with what little money Adobe gave us for training.

Therefore there should be payments for usage as well (as Shutterstock said it will comensate contributors).

5
Injustice for all, can you not see beyond your nose? I may be wrong but this is how I see it all unravelling in the near future...

AI is being used by the agencies to make contributors obsolete. Our images were being used to train AI and now we are being used to improve it by using AI. AI images will exponentially saturate the market and without retaining copyright to the new AI images, the agencies will remove us from the picture, keeping only the AI images which will dominate the market. New AI images will be genetated by customers at the prompt, adding more AI to the database. AI will be used to generate titles and tags. Soon they will not need human artists and photogtaphers at all.

Do not think so (even though I thought so as well in the past). Our role is changing (as well as for some other professions - AI is a tool for many of them, not a replacement). We are becoming the trainers of AI systems now at least in short and mid-term. There are some types of photographers which are going to be replaced more by AI (so their income from regular images will drop more but their income from AI training will increase) and vice versa.

6
would you sign an employment contract without knowing what you earn? Would you work for a month without knowing what would end up in your bank account? I maintain: no one does - but I may be off the mark.

I am not saying it is necesirally a good thing. All I say is that you cannot say it is bad neither according to the numbers you get because you do not know if you are getting less per sale or more. The reason is... you do not know how many images were sold using the system trained on your images (in comparison with ordinary ones), so you cannot even make an estimate, how much you should get per image. So, if you do not know, what number you should get, how can you say it is low or high? You also do not know how many other images of that topic were used in the system for that photo...

And that the "beneficial" could be, I do not see.

I do. It is a probability thing.

7
And the crumbs that we are given, are partially celebrated here! I dont get it!

I am just saying that you cannot say if these are crumbs or not because you do not know the numbers. I would say that this system can be even very beneficial for some type of photographers (for some topics).

8
We still hold the copyright, but they have taken an perpetual license to use our images and are paying us a one-time payment of an amazing $0.069 per image. At least that's the amount I got for each image. None of us would agree to something like that. But the way the agencies set it up, if you don't leave, you agree.

You have completely forgotten about one very, very, very important detail. Amount of sold images. 0,069 USD per image can be awful but also totally great amount of money. Let me explain, the following numbers are just for illustration of the concept.

Adobe will sell ordinary images for 1B USD and AI images (images created by AI trained on our images) for 10M USD per year. So, in our example case, they will sell 100-times more ordinary images than AI images. It is obvious that one cannot expect the same revenue from AI images as from the ordinary ones if they sell 100 times more. So, if your yearly revenue is... for example 10000 USD a year from 10000 images, it is totally fine to get anything above 1/100 of that (100 USD) per year from AI images.

So, the question is... do you really know what is the ratio between revenue from ordinary and AI images? I do not. Therefore, I cannot say if 0.069 USD per image is a lot or not. Can you?

9
I.e., say for 6 months you are at 10%. Then remainder of the year you are at 20%. The "average" rate is actually 15% (not 20%, which some ppl 'forget' about).

That is not entirely true. The fact is that the levels have different "size" (you jump very quickly in the beginning), therefore, you are in different levels for a different time period. In my case, I am at 15% approx. 3 weeks, than 5 weeks at 20 %, 8 weeks at 25 % and than remaining part of the year at 30 %. So, my average is around 27 %. Because of that, your average will be usually very close to the highest (or second highest) level achieved because you are going to spend the most time in it. Of course, there  are some exceptions, for example, if you barely achieve Level 2. And there is another catch, all these 10c sales which do not change with a change of level.

10
Shutterstock.com / Re: Blank image
« on: March 09, 2023, 17:00 »
This can happen if SS has deleted your image. The reason can be because of the trademark. Because my image was not suitable for commercial use anymore, it has dissapeared and was replaced by this blank stuff

11
OK does that make for better division of the target earnings?  :)

Thanks, that is much better and allows for deeper insights. Now, we can see that 40 % of all people are below 3 USD, which is not great. 60 % of all are below 10 USD and 75 % will fit below 20 USD.

I assume that these number will rise significantly next term as Shutterstock introduced the AI tool in general search.

12
Next time, it will be better if each range is larger than the previous one and not all of them are using 20USD steps. In such a case, there will be a large number of votes in one "basket" but nothing in the others, especially in the larger ones. 20 and 40 USD is 100% difference but 280 and 300 USD is just 7%. It will be miracle that someone will fit there. 0,01-20USD basket is extremely big and we have not got much information (is everybody near 0,01 or near 20? That is a huge difference). It will be better something like 0-1, 1-2, 2-4, 4-8, 8-16,... or slightly rounded (like 0-1, 0-2, 2-5, 5-10, 10-25, 25-50, 50-100, 100-250,...)

13
Shutterstock.com / Re: Shutterstock "Contributor Fund"
« on: December 17, 2022, 08:38 »
Will these royalties be ongoing for Ai images licensed where our work was used for machine learning?

It is understandable. These are the images used to train various AI systems or for generating images using AI and that can include hundreds of your images. I cannot imagine how to present such information in any reasonable way. Probably majority (if not all) of your images were used to train some AI system. Every image can contribute, so this can represent thousands of various sales.

14
... What I wanted to highlight is the fact that a contributor should be eligible for a commision every time when the image created with that topic is sold. ...

how would you decide from the generator phrase which images among millions were used in the training set?  eg "penguin in a top hat walking on mars'

once training is done, no images are actually accessed during the creation of AI images

That is a good question. You know what you have used for training each type of object (downloaded and fed to AI training system). I would assume it is going to be connected with keywords and training sets for them.

15
I won't say you're wrong, but from past experience, I'm not thinking this will be some kind of windfall for contributors.

I do not expect it either. What I wanted to highlight is the fact that a contributor should be eligible for a commision every time when the image created with that topic is sold. I expect more stable income (not necessarily larger income though).

16
I do not think it has to be so bad as it seems. I think our role will change from artists to providers of training images. I tried to elaborate more about the topic at https://artmino.com/shutterstock-and-ai-is-it-a-really-bad-deal/

17
Shutterstock.com / Re: Working together to lead the way with AI
« on: October 26, 2022, 08:07 »
  • What about users who don't have a dedicated IP but have a shared IP? The internet knows your IP address because it's assigned to your device and is required to browse the internet. Your IP address changes, though, every time you connect to a different Wi-Fi network or router.

This has nothing to do with IP address. IP is Intelectual Property, basically "content with author rights".

18
123RF / Re: Editorial vs. Commercial
« on: September 02, 2022, 09:15 »
I do not think it is possible now. Nevertheless, they will usually reject such images (for copyright, missing model release,...), so do not worry, these images will not get through. It is pity because 123RF is relatively reasonable agency for me.

19
Shutterstock.com / Re: SS strict rejection policy
« on: August 31, 2022, 03:19 »
I would say that resubmit always goes to human reviewer, but the first attempt... who knows. Basically all landscape images from full frame camera are rejected for focus (but they are just detailed - it seems that some AI does not understand grass straws and consider the details as noise). It is much easier to get accepted blurred images (soft corners) from a compact camera than perfectly sharp images from a full frame.

20
General Stock Discussion / Re: 1080 or 4k?
« on: March 23, 2022, 17:05 »
Hi mino216,
So you're saying HD is higher quality than 4k on the cell phone video?  I don't really know that much about video so I'm still pretty much learning the basics.

No, this is not what I have said :-) I usually take iPhone SE videos at 4K but because of some noise and compression artifacts, they have to be downscaled to Full HD in order to be accepted (Full HD video cteated by downscaling 4K video will be better than Full HD video taken at Full HD). It depends. Some scenes are OK in 4K and can be submitted as 4K, some are not and the quality is just not enough (it is "safer" to submit a nice Full HD created from 4K than to submit it as 4K where the quality standards and requirements are higher).

In other words, 4K quality from iPhone SE is often not enough for 4K quality requirements but it is more than enough to meet Full HD requirements (just downsample the not good enough 4K video to Full HD and you will have a very nice Full HD video).

21
General Stock Discussion / Re: 1080 or 4k?
« on: March 22, 2022, 04:07 »
The thing is that iPhone SE has some noise and compression issues so it would be wise to submit 4K clips from smartphone as Full HD and not 4K, so it does not matter much if DSLR or smartphone is used. At least, this applies for the first generation of iPhone SE which I use for stock video sometimes (I submit only minority of iPhone SE clips as 4K). I am not sure if the second and third generation are much better.

22
123RF / Re: Is it worth uploading vectors to 123RF ?
« on: February 18, 2022, 04:44 »
I have reasonable sales of vectors at 123RF but their review process is really painfully, painfully slow, especially for vectors (but it is not much better for photos either)

23
Adobe Stock / Re: Adobe Creative Cloud bonus codes for 2021?
« on: November 24, 2021, 06:56 »
I find that strange anyway. From my point of view, it is out of proportion. The jump from 150 to 5000 downloads is gigantic. And how many contributors really need all the products from the CC?

Why don't they offer intermediate Bonus solutions?
150 downloads - 1 product
500 downloads - 2 products
1500 downloads - 3 products
3000 downloads - 4 products
Just as an example.

I think that you can be happy that there is at least some option to get it basically for free. Adobe does not have to have such a program at all. If you are a photographer, you should be totally fine with 150 downloads for Photography plan which has almost all the things you may need. The tricky part is if you make also some illustrations and videos, though. But still, it is nice that Adobe offers that.

24
Shutterstock.com / Re: Focus Pocus?!
« on: November 04, 2021, 04:51 »
Almost every landscape photo that has trees with leaves is being rejected for focus (The main subject is out of focus or is not in focus due to camera shake). The other agencies have no issues with my photos. How is your experience?

The same here. Not so sharp landscape photos with weak corners from a 1-inch compact camera in 6 MPx are accepted almost all the time, sharper landscape images from full frame camera and corner-to-corner sharp lens are rejected in majority cases because of focus. Have to make them worse, blur them a little bit and upload in 6 MPx. Non-sense.

25
General Stock Discussion / Re: Dreamstime uploading not working?
« on: January 30, 2021, 15:57 »
Works in Chrome, does not work in Firefox.

Pages: [1] 2 3

Sponsors

Mega Bundle of 5,900+ Professional Lightroom Presets

Microstock Poll Results

Sponsors