IMatch 2025 Video and Release Notes Available as a Preview

[Mario]

I've finished the release notes for the upcoming major release of IMatch and made them public.

Check out the IMatch 2025 - What's New overview page, which has a video showcasing the highlights in IMatch 2025.

The release notes offer a detailed list of all new features, enhancements, changes and bug fixes :

https://www.photools.com/release-notes/

You'll find many requests made by you here in the community implemented in this version.

I've put a lot of work and effort into this new release. Countless hours. No sunshine. No food. No beer! Just work.

I'm sure every user will find something that allows him to work faster or easier or "do more" with this release. Sometimes even a small enhancements can save users hours of work or make IMatch work smoother.

Looking at the release notes, it's strange that a feature that required weeks or even months of work can be summarized in a single paragraph or two in the release notes

[mopperle]

Waiting for the final release.

[axel.hennig]

No beer!

Lucky you... healthy life! 

[Jingo]

I hope the release of 2025 allows you many beers and many hours of sunshine!

[Mario]

I hope the release of 2025 allows you many beers and many hours of sunshine!

Probably not. I prefer wine over beer and I live in Germany, which does not have that much sun in Winter. No sun reaches my cave anyway.

I'm looking forward to a bit more when the initial launch stress is over.

[axel.hennig]

I will definitely use this:

[Mario]

I will definitely use this:

I think this will be used by many users. It was requested, discussed and I've implemented it.

The tricky thing was that this directly affects the "face matching" when IMatch tries to assign persons to faces.
This is optimized to the hilt, the database system calling into my code to perform the vector math to match face fingerprints. Excluding persons on that level without ruining performance was a bit challenging.

But, like many things, this is faster in IMatch 2025 than it was ever before.

[loweskid]

Just out of interest, does AVX2 have any impact on IMatch?

I can't get Photoshop upgrades any more - it says I need AVX2 for the new AI features.  A bit annoying as I'm still paying the same subscription but I'll need to upgrade my computer next year as it won't install Win11 either.

[Mario]

IMatch requires only AVX, which is around for a decade (see hardware and software requirements). It's the oldest of the vector math extensions.
I could squeeze out a bit more performance with AVX2 or newer vector technology, but that's not that widely available. And building multiple IMatch versions for different processor technologies is not worth the overhead.

[ben]

Can't wait for the peek view 

[medgeek]

No beer!

Oh, no! Say it ain't so...

[kiwilink]

Can't Wait!!!  Thank you for all you do for us!!

[monochrome]

IMatch AI Tags
IMatch 2025 introduces unique AI Tags, which optionally allow you to separate AI-generated descriptions and keywords from human-made descriptions and keywords.

I can't order and install this version fast enough. Being able to describe and re-describe photos when better AI models come along without having to worry about data I have entered myself is just what I want.

When Imatch got the press-delete-in-viewer to cull images it boosted my workflow speed by 1000% - getting a good, functioning AI auto-tagger that doesn't get in the way will speed up organizing photos by the same order of magnitude at least.

I want this yesterday.

[Mario]

I can't order and install this version fast enough. Being able to describe and re-describe photos when better AI models come along without having to worry about data I have entered myself is just what I want.
 
When Imatch got the press-delete-in-viewer to cull images it boosted my workflow speed by 1000% - getting a good, functioning AI auto-tagger that doesn't get in the way will speed up organizing photos by the same order of magnitude at least.

Precisely. After working with various AI services and locally installed models for many months, I know that they are very good, but maybe not perfect - yet.

"Re-running" a set of images with different settings, a different prompt, different (newer) models or a different AI service is is perfectly possible and easy to do. If you can guide the AI with context information like "These images were taken in Paris" or "These are wedding photos of Maria's and Martins wedding", you get much better results. IMatch makes that easy.

The AI may not yet be perfect for all of your images. Or you'll learn that a newer AI in the future gets you better results. Or you come up with better prompts or your requirements change.

I've taken great care to make this very flexible but still easy to use.
AI capabilities change fast, and IMatch will automatically adapt to save you time and organize your files automatically.

The unique traits feature enables users to basically organize their files by any criteria they like.
Even users who don't care for an AI to generate descriptions or keywords might want to automatically organize files by criteria like lighting, animals, vehicles, indoor/outdoor, weather, mood, buildings or whatever.

I really don't know what users will make of this. This is why I've made this flexible and integrated variables to enable users  to reuse existing information in the database to support and guide the AI. This has proven to be very effective.

I'm sure IMatch users will do cool stuff with AutoTagger 2.0. I'm really looking forward to what users come up with.
And if a user does not see a need to use AI, that's also fine. There are lot's of other enhancements in IMatch 2025.

Or, as an AI (Ollama running on my notebook with the locally installed Llama3.2 model) would improve my rambling from above:

"I'd like to emphasize that after months of experience working with various AI services and locally installed models, I can attest that they're exceptionally good, but not yet perfect. 'Re-running' a set of images with different settings, prompts, newer models, or AI services is straightforward and easy to do. However, providing context information, such as 'These images were taken in Paris" or' "These are wedding photos of Maria's and Martin's wedding" significantly improves the results.

IMatch makes this process effortless. While the current AI capabilities may not be perfect for all your images, it's essential to remember that newer AI models can provide better results in the future, or you might discover more effective prompts or requirements changes. To accommodate these variations, I've designed IMatch with flexibility and ease of use in mind.

AI capabilities evolve rapidly, and IMatch will automatically adapt to save you time and organize your files efficiently. The unique traits feature enables users to categorize their files by any criteria they prefer, such as lighting, animals, vehicles, indoor/outdoor settings, weather conditions, mood, or building types.

Some users might not need AI-generated descriptions or keywords; however, the ability to automatically organize files by various criteria can still be beneficial. To support this flexibility, I've incorporated integrated variables that allow users to reuse existing information in the database, which has proven to be highly effective.

I'm confident that IMatch users will create innovative and interesting content with AutoTagger 2.0. I'm eager to see what creative ideas users come up with. Moreover, for those who don't require AI-driven features, there are numerous other enhancements in IMatch 2025 to explore."

I've just asked LLama to check for spelling and to improve the text I've wrote above.  And then copy & pasted the result here That's a free AI runner (Ollama) with a free model (LLama) doing all the work for me. What's not to like.