A good video. Ultimately in the early promos for this feature ON1 did NOT say "you will get better results by using the online service"; they said it was there as an alternative for users without a modenr GPU/computer. Once again, once the latest PR is released, they move the goalposts a little and now admit that actually, the online paid link is better. After they've no doubt processed all the pre-orders based on their misrepresentation. Indefensible. Fortunately after being stung on promises with 2023, I stuck with that and will ultimately go elsewhere when I replace it. Paying to edit photos in (still) a lower resolution isn't a direction I'm interested in. Scott, you are a very valuable resource in the online photography community, thanks.
Helpful as always Scott. The one thing no one seems to mention that I think is very relevant when considering using these tools is file size. Using either Gen AI tool creates an Onphoto file (needed so it remains non-destructive) which can be 200-300 MB in size. Throw in 4 or 5 AI masks (each of which creates a 200-300MB TIFF file) and you exploded your 50MB raw file into several GB. This is fine for a couple photos, but needs to be considered if you do it frequently.
You definitely want to keep an eye on file sizes in Windows Explorer or Mac Finder, as the case may be, especially, as you say when you use these tools on many images. When file size gets much above 100 MB (on my original 23 MB Raw file), I'm resizing the image. Photoshop's Image > Image Size (enter the original pixel count for the sensor, check Resample, set Preserve Details 2.0) is excellent at this. Resize AI in Photo RAW or Topaz Gigapixel AI are worth trying as well.
@@sounderdavis5446 The only thing I'm not sure about with this method is the huge TIFF files created for masks. I discovered it when my HD was full and it was all System Data (on Mac). Turns out almost all of it was TIFF files from creating masks. I don't think resizing the image has any impact on these files and they are HUGE.
I've had your issue as well: TIFF files with AI edits and masks can balloon up to 300-700 MB, even over a gig, yes. Have you tried resizing afterwards, comparing file sizes before-and-after while pixel peeping looking for loss of detail? A lot of what gets added in can be jettisoned in resizing especially if you're not making a big print. If you have access to Photoshop, try what I suggested. I can live with 118 MB but not >400 MB, generally. If you don't have PS, test alternates. I now always keep the original Raw file too in case technology finds something better (as it's been doing; did you delete slightly out-of-focus images because there was no fix 5+ years ago? OMG I sure did!)
For me any ONphoto files are only temporary files during the edit. Are you saying that you save every photo edit as an ONphoto file effectively duplicating each raw file with a partially developed demosaiced file and a final output file jpg or tif?
@@ianyorke2617 No I'm not saying I save every photo edit as Onphoto file. I'm saying users should be aware that they are being created using these tools, and the huge TIFF files are created (behind the scenes in system data) when creating any AI masks. The only thing ever mentioned is the sidecar files with edit instructions which are very small. I wasn't even aware of the system data files until they filled up my HD. ON1 support said they are considering allowing user to move the location of these files to external storage, but that's not available now.
Thanks so much for this video. It is very much needed. If ON1 discussed this at all, I missed it. I have had definite struggles with Generative Erase. I am using a Macbook Pro. I have only used the Local Gen AI. I have had zero good results in eliminating unwanted objects. Most often, new artifacts are added. Or, the results are smeared and useless. I have had some wonderful results with over exposed areas. I did a portrait shoot the other day and the sky was pretty brutal. I had numerous shots where parts of the sky were badly blown out. Turning down Whites and Highlights didn't produce satisfactory results nor did other editing choices. So, I tried Gen Erase. It worked great 95% of the time. For this purpose, it's a total asset. It feels like ON1 (which I love and have used for many years) have been disingenuous about these new features. Their videos on using Gen Erase and Gen Crop show amazing results with object removal and scene expansion that seem to take no time at all. The reality is far different. So thank you for jumping in with this necessary and clarifying information. It is very appreciated.
Hi Scott , very good informative video I agree with all comments posted here sounds like we were all led up the garden path very misleading information from on1 will never believe them again. VERY DISAPPOINTED.
As a current user of both ON1 Photo RAW and Adobe Photography Plan, for now, I'm going to continue to do this work in Adobe. These tools evolve gradually. Adobe has had a two-year head-start, and it shows here. Assuming that ON1 also continues to target programming resources to improving these AI tools, they too will get better at it and may at least catch up. But not yet.
Helpful explanation of the Gen options. To be fair to ON1 they have said that you may need to run the Local Gen options more than once and that each time you may get a slightly different result, each of which is saved as a separate layer, so you can then choose which one you prefer. I also suspect that ON1 will improve the AI Local Gen model over time.
This is a most opportune video. Thanks. In getting a stability AI account, there seem to be many different models to choose. For photo processing does it matter which model of stability ai you select.
Hi Scott. Thanks again for your excellent videos on various topics. Generative AI for removing and adding surfaces is a fairly new phenomenon, but will inevitably evolve further. Nice to see it also comes in ON1 Photo RAW, which I've been using for many years now and enjoy so much. I tested both removal and addition locally and via Stability. The result was better with Stability, but I wasn't completely satisfied with any of the options. "Dust removal" also doesn't seem to work in WIN 11 at the moment. As AI is increasingly used in photo editing softwares today, I ask myself the question: What level of GPU is needed today and tomorrow? Do you Scott, or anyone else have an opinion on that? I myself have a GeForce RTX 3060 Ti GPU 8Gb, which I've been quite happy with, but now I'm starting to think about something much sharper. For example. GeForce RTX 4070 Ti Super or even sharper. It would be interesting to hear what you think about this. It's not about cheap GPUs, after all.
Great video In the short time available, a complete photo (RAW) is never uploaded. You need about 1/3 to replace 2/3. My experience so far - Stability Ai is better when it comes to erase. with CROP I have so far found the results of local gen. AI better
Good example between the different AI models. To me Stability AI also could set a precedent for future content only available online [possibly for a cost] A slippery road I do not want to go down. But for professional use this may be an option in the tool kit that could make you money, quicker editing.
Professionals would require the AI generated content to be the same rez as the original file. This doesn't seem possible with this at this time. To me this is still very early days for the tech. Social media togs will perhaps lap it up though.
Scott, a very good video that hopefully most of the On1 users will look at rather than being upset without understanding what is really going on. One thing I haven't been able to determine is what is the resolution of Local AI and Stability AI. As you know, Adobe is limiting its solutions to 1024x1024 which as you have said and shown in this video does NOT give you results you can print. The same seems to be true with On1 but I don’t know the specifics, do you? In my opinion, AI is not ready for prime time which is ok as I have other choices which will often give me better results but with a lot more effort on my part.
Truly a "pig in a poke"". The video is very helpful, although it amazes me that you find the results acceptable. Generative AI in Oneone is not yet a serious editing tool. The stability AI service 'smells' like Adobe. I hope Onone will raise the performance of the local AI substantially in the 2025 version.
Another drawback on Stability AI I guess would be a large addition to the power consumption at those centres, where the AI processes are being done. Keeping the proces at my own machine I would guess does not use so much power as the cloud version. Wouldn’t you agree on that?
Hi Scott. Just to let you know that in your Pros and Cons slide, under the Local Gen AI heading there is a misspelling. Under Drawbacks, Potentailly should be Potentially, unless you wanted to highlight the ai 😃😃. HTH
Thanks for another great video. I just downloaded my copy of On1 photo raw max 2025. I am an On1 plus subscriber and have been since their first version that I pre ordered. In fact I have been using On1 products going back to 2014 when it was a plug in for Photoshop Elements. This helps me understand the difference between the two options. I clicked on the stability tab and it came up I had to subscribe or something to that effect. I backed out and tried local. Do you know if I set up an account with stability and pay $10.00, do my credits expire if I do not use them in a certain amount of time? I would not be needing it that often but it would be nice to have the option when I do need it.
Question - In your use of Local Gen AI, have you developed any way to 'repair the softness' issue using another tool/filter in Local Adjustment after using Local Gen AI?
I do not have personal experience with this ON1 feature, but I would first check to see that the "retouched" area is not low-resolution. Personally, I have been using Luminar for generative AI and have been very happy with it. It produces better results than Stablity AI at no extra charge. If you examine an image after repair, it looks perfectly natural (on my images). Today, Luminar Generative AI is much better than the original release a year ago.
Hello. My experience is bad in this sense. I have a good laptop, Intel i9, 32BG RAM, NVIDIA GeForce RTX 3070. The results using "local gen ai" are bad. So I think that the reason for two modules is that there is one, independent of the divice you own, the results will be bad and if you want good results you have to buy credits and use "Stability ai".
Hi Scott, I have to say I’m quite disappointed, and I feel that you’re not being entirely transparent. Comparing Stability AI to SmugMug is misleading. On one hand, you're promoting Generative AI as a built-in feature of the product, which is a reasonable expectation for a photo editing tool. However, you failed to mention a critical detail: this feature is completely unusable unless you have an NVIDIA or AMD GPU with at least 4GB of RAM-something you did not disclose in your video. I own a modern laptop with an Intel GPU, but unfortunately, I can't access the feature I paid for unless I subscribe to your hybrid solution. This was not communicated clearly, which has left me feeling frustrated. Moving forward, I will seriously reconsider my relationship with ON1. In Australia, this kind of omission could be considered misleading and deceptive conduct, and I believe more transparency is necessary to maintain trust with your customers.
I pre ordered 2025 based on videos presented by on1 on the new features coming. Turns out I ordered a "pig in a poke"
A good video. Ultimately in the early promos for this feature ON1 did NOT say "you will get better results by using the online service"; they said it was there as an alternative for users without a modenr GPU/computer. Once again, once the latest PR is released, they move the goalposts a little and now admit that actually, the online paid link is better. After they've no doubt processed all the pre-orders based on their misrepresentation. Indefensible.
Fortunately after being stung on promises with 2023, I stuck with that and will ultimately go elsewhere when I replace it. Paying to edit photos in (still) a lower resolution isn't a direction I'm interested in. Scott, you are a very valuable resource in the online photography community, thanks.
Helpful as always Scott. The one thing no one seems to mention that I think is very relevant when considering using these tools is file size. Using either Gen AI tool creates an Onphoto file (needed so it remains non-destructive) which can be 200-300 MB in size. Throw in 4 or 5 AI masks (each of which creates a 200-300MB TIFF file) and you exploded your 50MB raw file into several GB. This is fine for a couple photos, but needs to be considered if you do it frequently.
You definitely want to keep an eye on file sizes in Windows Explorer or Mac Finder, as the case may be, especially, as you say when you use these tools on many images. When file size gets much above 100 MB (on my original 23 MB Raw file), I'm resizing the image. Photoshop's Image > Image Size (enter the original pixel count for the sensor, check Resample, set Preserve Details 2.0) is excellent at this. Resize AI in Photo RAW or Topaz Gigapixel AI are worth trying as well.
@@sounderdavis5446 The only thing I'm not sure about with this method is the huge TIFF files created for masks. I discovered it when my HD was full and it was all System Data (on Mac). Turns out almost all of it was TIFF files from creating masks. I don't think resizing the image has any impact on these files and they are HUGE.
I've had your issue as well: TIFF files with AI edits and masks can balloon up to 300-700 MB, even over a gig, yes. Have you tried resizing afterwards, comparing file sizes before-and-after while pixel peeping looking for loss of detail? A lot of what gets added in can be jettisoned in resizing especially if you're not making a big print. If you have access to Photoshop, try what I suggested. I can live with 118 MB but not >400 MB, generally. If you don't have PS, test alternates. I now always keep the original Raw file too in case technology finds something better (as it's been doing; did you delete slightly out-of-focus images because there was no fix 5+ years ago? OMG I sure did!)
For me any ONphoto files are only temporary files during the edit. Are you saying that you save every photo edit as an ONphoto file effectively duplicating each raw file with a partially developed demosaiced file and a final output file jpg or tif?
@@ianyorke2617 No I'm not saying I save every photo edit as Onphoto file. I'm saying users should be aware that they are being created using these tools, and the huge TIFF files are created (behind the scenes in system data) when creating any AI masks. The only thing ever mentioned is the sidecar files with edit instructions which are very small. I wasn't even aware of the system data files until they filled up my HD. ON1 support said they are considering allowing user to move the location of these files to external storage, but that's not available now.
Thanks so much for this video. It is very much needed. If ON1 discussed this at all, I missed it. I have had definite struggles with Generative Erase. I am using a Macbook Pro. I have only used the Local Gen AI. I have had zero good results in eliminating unwanted objects. Most often, new artifacts are added. Or, the results are smeared and useless.
I have had some wonderful results with over exposed areas. I did a portrait shoot the other day and the sky was pretty brutal. I had numerous shots where parts of the sky were badly blown out. Turning down Whites and Highlights didn't produce satisfactory results nor did other editing choices. So, I tried Gen Erase. It worked great 95% of the time. For this purpose, it's a total asset.
It feels like ON1 (which I love and have used for many years) have been disingenuous about these new features. Their videos on using Gen Erase and Gen Crop show amazing results with object removal and scene expansion that seem to take no time at all. The reality is far different.
So thank you for jumping in with this necessary and clarifying information. It is very appreciated.
Hi Scott , very good informative video I agree with all comments posted here sounds like we were all led up the garden path very misleading information from on1 will never believe them again. VERY DISAPPOINTED.
Clearly presented and excellent information: thank you a million, Scott!
Glad it was helpful!
As a current user of both ON1 Photo RAW and Adobe Photography Plan, for now, I'm going to continue to do this work in Adobe. These tools evolve gradually. Adobe has had a two-year head-start, and it shows here. Assuming that ON1 also continues to target programming resources to improving these AI tools, they too will get better at it and may at least catch up. But not yet.
Excellent explanation plus making making photographers aware of the erase feature differences... Thanks Scott :)
Very useful Scott - as your videos are. Food for thought. Thanks
Glad to help
Helpful explanation of the Gen options. To be fair to ON1 they have said that you may need to run the Local Gen options more than once and that each time you may get a slightly different result, each of which is saved as a separate layer, so you can then choose which one you prefer. I also suspect that ON1 will improve the AI Local Gen model over time.
I won't be using Stability AI. Thank you Scott for this very important information.
Great video. thanks I think I will only play around with the local one.
Thanks Scott for making things clear.For my taste this feature is useless as it looks now.
This is a most opportune video. Thanks. In getting a stability AI account, there seem to be many different models to choose. For photo processing does it matter which model of stability ai you select.
Hi Scott.
Thanks again for your excellent videos on various topics.
Generative AI for removing and adding surfaces is a fairly new phenomenon, but will inevitably evolve further. Nice to see it also comes in ON1 Photo RAW, which I've been using for many years now and enjoy so much.
I tested both removal and addition locally and via Stability. The result was better with Stability, but I wasn't completely satisfied with any of the options. "Dust removal" also doesn't seem to work in WIN 11 at the moment.
As AI is increasingly used in photo editing softwares today, I ask myself the question: What level of GPU is needed today and tomorrow?
Do you Scott, or anyone else have an opinion on that? I myself have a GeForce RTX 3060 Ti GPU 8Gb, which I've been quite happy with, but now I'm starting to think about something much sharper. For example. GeForce RTX 4070 Ti Super or even sharper.
It would be interesting to hear what you think about this. It's not about cheap GPUs, after all.
Great video In the short time available, a complete photo (RAW) is never uploaded. You need about 1/3 to replace 2/3. My experience so far - Stability Ai is better when it comes to erase. with CROP I have so far found the results of local gen. AI better
Thanks for sharing
Good example between the different AI models.
To me Stability AI also could set a precedent for future content only available online [possibly for a cost] A slippery road I do not want to go down.
But for professional use this may be an option in the tool kit that could make you money, quicker editing.
Professionals would require the AI generated content to be the same rez as the original file. This doesn't seem possible with this at this time. To me this is still very early days for the tech. Social media togs will perhaps lap it up though.
Scott, a very good video that hopefully most of the On1 users will look at rather than being upset without understanding what is really going on. One thing I haven't been able to determine is what is the resolution of Local AI and Stability AI. As you know, Adobe is limiting its solutions to 1024x1024 which as you have said and shown in this video does NOT give you results you can print. The same seems to be true with On1 but I don’t know the specifics, do you? In my opinion, AI is not ready for prime time which is ok as I have other choices which will often give me better results but with a lot more effort on my part.
Truly a "pig in a poke"". The video is very helpful, although it amazes me that you find the results acceptable. Generative AI in Oneone is not yet a serious editing tool. The stability AI service 'smells' like Adobe. I hope Onone will raise the performance of the local AI substantially in the 2025 version.
Another drawback on Stability AI I guess would be a large addition to the power consumption at those centres, where the AI processes are being done. Keeping the proces at my own machine I would guess does not use so much power as the cloud version. Wouldn’t you agree on that?
Hi Scott. Just to let you know that in your Pros and Cons slide, under the Local Gen AI heading there is a misspelling.
Under Drawbacks, Potentailly should be Potentially, unless you wanted to highlight the ai 😃😃.
HTH
Ha! Maybe I should have let the AI write the slide :)
Thanks for another great video. I just downloaded my copy of On1 photo raw max 2025. I am an On1 plus subscriber and have been since their first version that I pre ordered. In fact I have been using On1 products going back to 2014 when it was a plug in for Photoshop Elements. This helps me understand the difference between the two options. I clicked on the stability tab and it came up I had to subscribe or something to that effect. I backed out and tried local. Do you know if I set up an account with stability and pay $10.00, do my credits expire if I do not use them in a certain amount of time? I would not be needing it that often but it would be nice to have the option when I do need it.
Stability ai has terms of agreement that appears to give them rights to your photo.
Question - In your use of Local Gen AI, have you developed any way to 'repair the softness' issue using another tool/filter in Local Adjustment after using Local Gen AI?
I do not have personal experience with this ON1 feature, but I would first check to see that the "retouched" area is not low-resolution. Personally, I have been using Luminar for generative AI and have been very happy with it. It produces better results than Stablity AI at no extra charge. If you examine an image after repair, it looks perfectly natural (on my images). Today, Luminar Generative AI is much better than the original release a year ago.
Hello.
My experience is bad in this sense.
I have a good laptop, Intel i9, 32BG RAM, NVIDIA GeForce RTX 3070.
The results using "local gen ai" are bad. So I think that the reason for two modules is that there is one, independent of the divice you own, the results will be bad and if you want good results you have to buy credits and use "Stability ai".
HiScott. I download and installed the trial version ON1 Photo Raw 2025, I can not find the the Generative Eraser icon... Thanks for your video
Click on the two circle icon just above the fingerprint icon and move your mouse to the top and the gen eraser tool will appear.
Hi Scott,
I have to say I’m quite disappointed, and I feel that you’re not being entirely transparent. Comparing Stability AI to SmugMug is misleading. On one hand, you're promoting Generative AI as a built-in feature of the product, which is a reasonable expectation for a photo editing tool. However, you failed to mention a critical detail: this feature is completely unusable unless you have an NVIDIA or AMD GPU with at least 4GB of RAM-something you did not disclose in your video.
I own a modern laptop with an Intel GPU, but unfortunately, I can't access the feature I paid for unless I subscribe to your hybrid solution. This was not communicated clearly, which has left me feeling frustrated.
Moving forward, I will seriously reconsider my relationship with ON1. In Australia, this kind of omission could be considered misleading and deceptive conduct, and I believe more transparency is necessary to maintain trust with your customers.
Why you sound like speaking through phone😮
It's a gift. Unless you're watching on your phone in which case I am speaking through a phone.