Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

What You Can and Can't Make With iOS 18.2's Genmoji Feature

Expos of 1969 said:
Siri - "How does Ms. Gemini do that?" Why can't I do that?" "Mr. Cook, why, why?"
Click to expand...
Private Cloud Compute will open up an Apple solution for the specific issue of "not enough device-local compute, but Apple still wants to retain privacy".

The extent to which it solves the problem (real or perceived) of "Apple models, even running at scale, are not as good as what I want" is unclear.
The extent to which third parties will be willing to use that Apple Infrastructure (especially the cases, like Claude or ChatGPT, that people want to use) is also unclear.
 
Both Image Playground and Genmoji are incredibly underwhelming. It's not even the limiting factors that are the problem. It's the poor art model they are using. It's really weak. Cute is fine. It's just poor artistic aesthetics. Re-train it on some better artwork.
 
  • Love
  • Like
Reactions: 9081094 and turbineseaplane
Playing with genmoji gets my 15 pro really hot, pretty quickly. (Goes hand-in-hand with the battery drain of course!)
 
I’m hoping it’ll have improved by the time we get it in the EU in April. I’m a frequent user of most iMessage visual features — Bubble/Screen effects and the newer Text Effects — and would use both Genmoji and Image Playground regularly.
(Am I hoping in vain?)
 
alchemistmuffin said:
This is why Genmoji is severely limited and why some prompt doesn’t spit out result: this is all done 100% on device. This is what makes Apple Intelligence different from other Generative AI: it’s all on device generation. And unfortunately the neural engine on the M series and A series chip is still about years away from being able to have generative ability that ChatGPT or Dall-E can do via cloud AI.
Click to expand...

Yes and no.

Apple would have a hard time getting Dall-E quality work on device, unless your device were a large memory Mac Studio. That said, in another thread someone turned me on to Draw Things, a free AI image generator that allows you to download models from the web. I've been using it on the Mac for a couple of weeks. It makes amazing images. There is a learning curve and it isn't for people looking for Image Playground style simplicity. It runs on iPhone also. The point of mentioning this is that it is possible to generate high-quality images using an M or A chip. It isn't fast for large high-quality images, but it can be done.

The following image was generated using Draw Things on an iPhone 16 Pro Max. The prompt was "Cowboy riding a T-Rex through a jungle." It took approximately 4 minutes and 3% of my battery. I was able to generate a similar picture using Draw Things program on a Mac Studio M2-Ultra in approximately 20 seconds.
 

Attachments

  • IMG_0262.PNG
    IMG_0262.PNG
    1.9 MB · Views: 44
Last edited:
  • Like
Reactions: name99
rjjacobson said:
I am finding even more issues with Apple Intelligence given so much is done on the powerful chips on our devices. My battery usage per hour doubled under 18.2 betas (except beta 5 and RC 1) for both my M4 iPad Pro and my 15 PM.
Click to expand...
As you noted, the problem with your battery usage was on beta versions, not on RC 1. You should always expect that beta releases have debugging code turned on and so will run slower (and take more battery) than released version.
 
vantelimus said:
As you noted, the problem with your battery usage was on beta versions, not on RC 1. You should always expect that beta releases have debugging code turned on and so will run slower (and take more battery) than released version.
Click to expand...
No the issue came back with RC 2 and also the official release of 18.2 ipadOS so not just beta it continued after the beta for me on my M4 iPad Pro maybe I was not clear where it occurred? Ah i missed saying official release of 18.2 my bad but earlier in this thread I did say that for me only 18.2 beta 5 and RC 1 gave me the 5 percent per hour. The official release of 18.2 was back at the 10 percent per hour. So something was turned back on in the release version also for me not sure what wonder if the debugging code was still on. All I know turning off Apple intelligence got me back to normal battery drain
 
jarman92 said:
You can, just type in their name then tap the little card that pops up.
Click to expand...
I don't have everyone tagged by name in my camera roll. And for the ones I do when you tap on them in the sheet for Genmoji it only presents me with 5 or so entries that the system has pre-selected. There's no way for me to pick another photo of them as a source.

It's way too restrictive considering that Image Playground lets you just select any photo of any person from your camera roll.
 
MisterSavage said:
I don't have everyone tagged by name in my camera roll. And for the ones I do when you tap on them in the sheet for Genmoji it only presents me with 5 or so entries that the system has pre-selected. There's no way for me to pick another photo of them as a source.

It's way too restrictive considering that Image Playground lets you just select any photo of any person from your camera roll.
Click to expand...

Works for me, I can just tap “Choose Other Photo…” and pick any photo I want.

Also, you actually can use a generic person—tap the name, then “Emoji” when the “Choose a Person” screen pops up.
 

Attachments

  • IMG_0848.png
    IMG_0848.png
    305.9 KB · Views: 40
Apple should avoid the AI business because their overly controlled approach would make it a restrictive, walled garden. Instead, they should focus on offering curated experiences with clear limitations. For example, instead of pretending you can create your own emojis, they could provide thousands of pre-approved options that users can combine freely, ensuring no misuse but still allowing some creativity. It’s not AI or true freedom, but at least it wouldn’t frustrate users with useless features. If Apple embraced a curated yet flexible system, it might work, but a lazy, overly restrictive approach would just feel like a gimmick designed to market their devices to younger audiences. If they can’t offer real value, they shouldn’t bother.
 
  • Like
  • Disagree
Reactions: vantelimus, rjjacobson and Skimac
I can honestly say now that I’m glad Apple isn’t introducing it in Europe. What a complete disaster and waste of software engineers and money. Won’t be long and Apple is following the route of Atari. Making only toys for the young generation.
 
  • Like
Reactions: Skimac
jarman92 said:
Works for me, I can just tap “Choose Other Photo…” and pick any photo I want.

Also, you actually can use a generic person—tap the name, then “Emoji” when the “Choose a Person” screen pops up.
Click to expand...

For me "choose other photo" in Genmoji opens another sheet with a small selection of pre-selected photos. It does not open the photo library where I can select any photo.

I didn't want a generic person. I have wanted to use people that are not in the small selection of people available as a choice. Something I can easily to in Image Playground with "choose photo", which opens my entire photo library.
 
TechTeacher101 said:
You guys are out here trying to make some next level stuff. All I wanted to make was one based on a turkey for thanksgiving. Nope. Try it. Doesn't work. Santa does though.
Click to expand...
Not acceptable?
I do agree the language model is too dumb, requires too much futzing to get something it accepts; and, as I have said before, I don't get this because Apple's generic language model is much more powerful.
Screenshot 2024-12-15 at 1.53.05 PM.jpeg
 
It's so disappointing to be on iPhone 14 and not be able to waste my battery on this stuff. AVP and Genmoji. What a legacy for Tim.
 
This is a really good point that not many folks have touched on. The fact is that all of this generative AI stuff is a huge energy drain— it has to come from somewhere. When it’s on our devices, we notice; when it’s on the cloud, this becomes way less visible. But Microsoft reopened a nuclear plant that’s been closed for years in Pennsylvania to fuel this. Tons of other, more polluting energy is firing up for “genAI”. And for what? Bad cartoons? Low value entertainment? While emissions go up and up? This sort of consumer focused generative AI isn’t the future by any means and in the present it has real effects that are pulling us backwards not forwards on many fronts.

rjjacobson said:
I am finding even more issues with Apple Intelligence given so much is done on the powerful chips on our devices. My battery usage per hour doubled under 18.2 betas (except beta 5 and RC 1) for both my M4 iPad Pro and my 15 PM. That meant for example on my M4 iPad Pro I went from the normal 5 percent per hour I had with it when I got it in May to 10 percent per hour. iPhone was similar draining. So all the other aspects of Apple Intelligence that are done locally vs offline server generated use too much power in my experience for the limited value of Apples Highly restricted and guardrailed AI implementation. I have for now turned it off on all my apple devices (M4 iPad Pro, 15 PM, M2 16 inch MacBook Pro). Just not sure there is much there i need. I do my own research anyway as a geologist and I do not need Apple devices to substitute for my brain. In fact at 74 it is good to use my brain rather than let my devices take over it keeps my mind stimulated. No thanks apple no Apple intelligence for now for me unless I can find a worthwhile use
Click to expand...
 
  • Haha
  • Like
Reactions: ipedro and rjjacobson
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back