.. there's a free AI image creator at https://www.bing.com/images/create
You need a Microsoft account to login.
This is what I got for "a cat piloting a starship"
(https://i.ibb.co/sJPJMZr/catpilot.jpg)
https://www.bing.com/images/create/an-army-of-tanks-driven-by-mice/1-657c73d1a8534588821da863720c2c64?id=9485rXe45vjWssHD1vZnTA%3d%3d&view=detailv2&idpp=genimg&FORM=GCRIDP
https://www.bing.com/images/create/a-pyramid-shaped-face-man-shouting/1-657ca58c82f4457c8bb91fd5d297271a?id=nZPkHCz7u6%2beWndpwNhMgw%3d%3d&view=detailv2&idpp=genimg&FORM=GCRIDP
I started one over 4 hours ago and it's still "creating". I could have knocked something up myself in that time.
something must have gone wrong there, should take about three mins.
https://www.bing.com/images/create/goldfish-dressed-as-a-porcupine-shakes-the-hand-of/1-657cbf1ad88446cfbb48d10f84587a9d?id=SRBxe1osQj0W8cLaOKFEUQ%3d%3d&view=detailv2&idpp=genimg&FORM=GCRIDP&mode=overlay
Have to say it's impressive. And fun. But I asked for a library in the style of Escher and that looked really good!
It finished at last - "A Dachshund eating a banana":
(https://i.imgur.com/GX1HTMW.jpg)
Don't ask for a cat dressed as adolf hitler Playing darts. it doesn't like it
I'm starting a (very modest, part time) Internet radio station called Ravenstone Radio, and it came up with some pretty good promo artwork:
(https://i.ibb.co/Jkx09C7/rrstudio.jpg)
(https://i.ibb.co/kq3dCFW/rradio-1.jpg)
I was really impressed with this one, though I won't use it:
(https://i.ibb.co/VtPWtYy/dj-at-rr.jpg)
Quote from: Slim on December 15, 2023, 09:20:12 PMI was really impressed with this one, though I won't use it:
(https://i.ibb.co/VtPWtYy/dj-at-rr.jpg)
You've seem to have reversed your male pattern baldness
;)
Well I admit my hairline has receded slightly but "baldness" is overstating it
Kerry King eating Krabs
(https://i.ibb.co/fq8RFbJ/KK-Crabs.jpg)
https://www.bing.com/images/create/a-planet-earth-that-has-lost-its-soul-and-hope-for/1-657dbf145f8045f8b5a8d40b5f51a58f?id=kFNXoErPFtqGuYAdTSNTJA%3d%3d&view=detailv2&idpp=genimg&FORM=GCRIDP
(https://i.ibb.co/LgzjB4d/e1248a07-c047-474e-adc8-c77266763f09.jpg)
Rockband Rush playing hide & seek
Bit weird, but OK. Did a bunch of others I actually liked.
Library Escher style
Like it a lot
(https://i.ibb.co/FmwJRpc/a03cf62d-645c-4cb8-a286-51a97291b7c1.jpg)
Chihuahua nurse on mountain top
(https://i.ibb.co/WFvwYyk/1785cd3a-bc84-4d06-91c0-b52fd270e7c0.jpg)
White horse firefighter
(https://i.ibb.co/qxHfFbz/a807666c-7128-42bd-a77d-6ad4a7eff3bd.jpg)
This one I really enjoyed:
Graphic illustration stamp of Rotterdam in Winter
It's funny because the AI clearly has no clue of landmark placements. It's all wrong, but also it's fun to look at.
(https://i.ibb.co/1f7Dhqw/1876284a-3b73-4017-a6ed-6b9df5ac8b1f.jpg)
Sorry folks, just enjoying myself here...
Jugendstil map of New York
(https://i.ibb.co/6BJt6K7/0325a6f3-00f3-4dd1-a46b-4ef8c19baf51.jpg)
And wanted a self portrait:
Epic Warrior with Covid
No clue what's going on with the St. George's though I did not specify that, nor the style.
(https://i.ibb.co/LzSDPmJ/aac0aa5b-b948-47c0-a0c5-84027f2b6d8f.jpg)
Think this is pretty self explanatory:
(https://i.ibb.co/0yPKpyF/Cat.jpg)
50s style Postcard of a motel on mars
(https://i.ibb.co/qrs5Nhh/36398e3f-c4b2-4fc8-be32-a0c8657be51d.jpg)
Medusa Octopus
(https://i.ibb.co/17Fs2M4/f3659393-93e8-4383-bfe5-704b419b56c5.jpg)
Greek philosopher dressed as a basketball player Polaroid photo style
(https://i.ibb.co/mRYZFy1/e690ac85-5cdf-4e77-889a-4d97bea751c4.jpg)
For Christmas I got my son tickets to see Irvine Welsh when he comes to Dublin next year. Will give this to him on Christmas Day
https://www.bing.com/images/create/tickets-to-see-irvine-welsh-the-scottish-author-wh/1-65859176546c433eb588dcb7cfc45bb0?id=dT6eUjxMnylWKJYdQynKpQ%3d%3d&view=detailv2&idpp=genimg&FORM=GCRIDP&mode=overlay
And Louise tickets to see The Mousetrap
https://www.bing.com/images/create/a-mouse-trap-next-to-tickets-to-see-the-mousetrap-/1-65858f211223495b954281113fd95e81?id=KhpZGghAinSA7pa3%2bFmlFg%3d%3d&view=detailv2&idpp=genimg&FORM=GCRIDP&ajaxhist=0&ajaxserp=0
Some dude at reddit managed to create/recreate Haken album covers. He uses very descriptive sentences to generate these (look down in the comments to see how he did it). Very clever indeed:
https://www.reddit.com/r/Haken/comments/18ocffe/budget_haken_ai_generated_album_covers/
Did something similar for the Rainbow debut album, unfortunately it's dropped off my list of recent images now. Just had another go - this one's not bad.
(https://i.ibb.co/yXkzF1z/rainbowdebutai.jpg)
Take the Real or AI image quiz:
https://www.bbc.co.uk/bitesize/articles/zg2d6rd
I got 4/8, which is roughly what you'd expect from tossing a coin. Only one of them looked obviously AI to me and even that wasn't bad.
Got 4 as well
I've had a few goes now just to experiment, but any text I use gets corrupted in the image, translated with extra or different letters. What gives?
Here is an experiment where I was trying to get the AI to generate an image to represent the "Scandinavian Monkey Business" tour:
(https://th.bing.com/th/id/OIG.s1H2Jas.U9d2QM7DRoHi?pid=ImgGn)
Here is another where "Rosalind Woods Collector" has ended up with way too many LLLs and OOOs.
(https://th.bing.com/th/id/OIG.gt12ToDoTWJCb2KrIQKL?pid=ImgGn)
Yes it's odd that text generation is so unreliable. You'd think that it would be the simplest aspect of the task. But it's a preview version so hopefully it will get ironed out.
This is really a bit trippy
https://www.krea.ai/apps/image/realtime
Create images from prompts - the images build up and transform as you type. Then you can modify them by dragging shapes around in the left pane.
Wasn't happy with the composition of this image, not enough space at the top above the spire.
(http://bikediary.uk/images/241005naseby.jpg)
So I tried a free AI image "uncrop" tool online, https://www.insmind.com/uncrop-ai/
It worked nicely. The version you get when using the site for free is nastily compressed as you can see, but if you paid for an account you'd get a decent copy.
(https://i.ibb.co/sjR8spz/nase-extend-compress.jpg)
Ahem, you haven't got all of your bike in btw
;D ;)
This is a very capable image generator:
https://theresanaiforthat.com/@aldous/imaginator-99
I tried "a Japanese woman kneels to put clothes in her washing machine. She's wearing a white shirt and a red skirt. In the background, through her window, we see a rocket taking off".
(https://i.ibb.co/y6n0hSz/imaginator-ai.jpg)
It's got the shadow from the window just right. Very natural image. I'd have been more impressed if the washing machine door had been open.
So I tried the same prompt with the generator at bing.com, and I got four anime images. I added "photorealistic" to the same prompt and got two images. This is the better one - much less natural than imaginator-99 above.
I'm not sure about the bonsai tree coming out of the skirting board, her right leg looks overlong, the washing machine door doesn't quite look right. But at least it's open.
(https://i.ibb.co/wcScv1T/bing-wm.jpg)
(https://i.ibb.co/z5wcsk9/ai-image2-250106.webp)
(https://i.ibb.co/GJZFnxj/ai-image250106.webp)
Scary, isn't it? Pretty soon you won't know if anything you see on a TV, phone or computer screen is real.
There are already thousands of AI porn images out there that look very real. But think of the implications for businesses that market clothing, for example. Photoshoots will be a thing of the past. There'll be no need to hire a model or a photographer, or organise a location.
I asked ChatGPT for:
An extremely unremarkable iPhone photo with no clear subject or framing—just a careless snapshot. The photo has a touch of motion blur, and is mildly overexposed from uneven sunlight. The angle is awkward, the composition nonexistent, and the overall effect is aggressively mediocre—like a photo taken by accident while pulling the phone out of a pocket to take the selfie. It's of Geddy Lee playing oldschool minecraft on an old computer in a bedroom. he's looking over his shoulder at the camera. Candid.
(https://i.ibb.co/Z6T205jC/geddy-sora.jpg)
Computer in a bedroom taken very literally.
(https://i.ibb.co/v6mH2z4J/alex-toaster.jpg)
Looks like Geddy's image should have had porn on the PC ;D
Alex is definitely cultivating a bit of a 'whisky nose', like Neil's but not quite as well developed. May have been 'created' by AI but I think not :)
That image also looks 'sans hairpiece' so definitely candid ;D
Quote from: David L on April 15, 2025, 02:32:08 PMAlex is definitely cultivating a bit of a 'whisky nose', like Neil's but not quite as well developed. May have been 'created' by AI but I think not :)
Not sure what you mean? I asked ChatGPT for an image of Alex Lifeson repairing a toaster in a garage, looking over his shoulder to the camera.
Quote from: Slim on April 15, 2025, 03:36:56 PMQuote from: David L on April 15, 2025, 02:32:08 PMAlex is definitely cultivating a bit of a 'whisky nose', like Neil's but not quite as well developed. May have been 'created' by AI but I think not :)
Not sure what you mean? I asked ChatGPT for an image of Alex Lifeson repairing a toaster in a garage, looking over his shoulder to the camera.
I mean:
If you asked AI to create a 'current' image of Alex, it may take into account his propensity to indulge quite heavily in alcohol and give his nose a bit of a purplish bloom. Something akin to 'The Professor's' whisky-induced hooter.
Or it may have just used a stock image, because he's definitely got the start of a pint-swiller's proboscis. Possibly the fault of Henderson's brewery or hanging at Geddy's.
It does use stock images in a sense, because it will have been trained on multiple images of him. But it's unlikely that it's a simple incorporation of a single stock image. I think. Actually I'll ask it ..
QuoteQuoteThe image you provided of Alex Lifeson earlier. It was quite realistic. Is the particular representation of Alex' head possibly derived from a stock image? Or is it a reimagination based on your training data, which (presumably) contains multiple images of him - so you constructed a new image based on information points from those?
Great question—and you're right on both counts in a way, but here's the breakdown:
The image I generated of Alex Lifeson isn't taken from a specific stock photo or an existing image. Instead, it's a synthetic reconstruction—a reimagining based on a large number of visual and textual data points from training. The model has been trained on many images of public figures like Lifeson, including how he looks at various ages, his general facial features, hair, build, and common expressions. So when you ask for a realistic image of "modern-day Alex Lifeson," the model draws from that overall pattern and creates something new that resembles him in a plausible, coherent way—kind of like a digital impressionist who's seen a bunch of photos but isn't copying any one in particular.
It's a bit like what a sketch artist might do if they'd studied someone's face from different angles and then drew them doing something the artist had never seen before.
The result might feel a bit uncanny at times—realistic, but not quite identical to a specific known photo. If you put the generated image next to real photos, you'd likely spot differences, even though the vibe and features might be spot-on.
Scary!
Quote from: Slim on April 15, 2025, 06:26:23 PMIt does use stock images in a sense, because it will have been trained on multiple images of him. But it's unlikely that it's a simple incorporation of a single stock image. I think. Actually I'll ask it ..
QuoteQuoteThe image you provided of Alex Lifeson earlier. It was quite realistic. Is the particular representation of Alex' head possibly derived from a stock image? Or is it a reimagination based on your training data, which (presumably) contains multiple images of him - so you constructed a new image based on information points from those?
Great question—and you're right on both counts in a way, but here's the breakdown:
The image I generated of Alex Lifeson isn't taken from a specific stock photo or an existing image. Instead, it's a synthetic reconstruction—a reimagining based on a large number of visual and textual data points from training. The model has been trained on many images of public figures like Lifeson, including how he looks at various ages, his general facial features, hair, build, and common expressions. So when you ask for a realistic image of "modern-day Alex Lifeson," the model draws from that overall pattern and creates something new that resembles him in a plausible, coherent way—kind of like a digital impressionist who's seen a bunch of photos but isn't copying any one in particular.
It's a bit like what a sketch artist might do if they'd studied someone's face from different angles and then drew them doing something the artist had never seen before.
The result might feel a bit uncanny at times—realistic, but not quite identical to a specific known photo. If you put the generated image next to real photos, you'd likely spot differences, even though the vibe and features might be spot-on.
Interesting response. The largest proportion of images that it referred to must've been without the syrup because it seems to have 'aggregated' those without and those with and came up with an image that looks far more like he's sporting his own hair rather than the full rug. I think it's definitely a bug with AI that needs to be ironed out (the 'rug' bug?).
Try Ian Paice dancing with a Ladyboy in a laundrette or, better still, Richie Blackmore playing mousetrap with the Pope. We'd know for sure if the 'rug' bug needs fixing.
(https://i.ibb.co/RTbgzZhz/geddy-cycling.jpg)
Quote from: Slim on May 26, 2025, 10:58:02 PM(https://i.ibb.co/RTbgzZhz/geddy-cycling.jpg)
AI obviously forgot about the cool Chopper gear change lever. Mine had two as it was a 5-speed, which was quite rare. Now that my memory has been jogged I've remembered that Raleigh also produced a drop handlebars racing version. I don't remember ever seeing one other than in Dales Cycles in their original Glasgow store. Great bike shop Dales. Bought my e-bike there last year. Really knowledgeable, friendly staff and still family owned and run. I was quite surprised at their stock levels on my last visit. Must be a huge financial commitment.
Quote from: Slim on January 10, 2024, 01:51:51 PMYes it's odd that text generation is so unreliable. You'd think that it would be the simplest aspect of the task. But it's a preview version so hopefully it will get ironed out.
I tried a few text-based images with Gemini last night and noticed that this was much improved:
(https://i.ibb.co/WNwWBbDR/ravoradio.jpg)
(https://i.ibb.co/pvQPmVT8/middleclass.jpg)
I pointed this out to the AI, which replied:
(https://i.ibb.co/QjK029NH/obser.png)
.. however I've just tried the bing.com image generator, and it still messes up text.
(https://i.postimg.cc/mcMcrB1H/IMG-4223.png) (https://postimg.cc/mcMcrB1H)
Thought I'd give co pilot a go..
Quote from: Fishy on May 31, 2025, 08:12:35 PM(https://i.postimg.cc/mcMcrB1H/IMG-4223.png) (https://postimg.cc/mcMcrB1H)
Thought I'd give co pilot a go..
Alex before his stomach stitch up then
(https://i.ibb.co/nN2XwPQB/sunbguitar.webp)
https://huggingface.co/spaces/multimodalart/Qwen-Image-Edit-Fast
"Make the guitar gloss black"
(https://i.ibb.co/GQxFYbcd/blackguitar.webp)
So now you know you should stick with the burst.
"Remove the scratch plate" ;)
I suspect that would work actually. I wasn't necessarily expecting it to recolour the "poker chip" or the knobs and scratchplate, but .it does look exactly like a Studio Les Paul. The only wrinkle is the text on the poker chip.
It did change the colour of the reflection on the lower pick up.
Quote from: Thenop on August 31, 2025, 12:45:30 PMIt did change the colour of the reflection on the lower pick up.
Yes I thought that was odd .. it redid the text on the poker chip as well, badly.
I asked the same AI to furnish a photo from when I was clean shaven with a full beard. Don't have nightmares.
(https://i.ibb.co/60858193/fullbeard.webp)
Then because I've never had dark body hair anywhere, I got Gemini to make it go grey. Can't helping feeling there's a bit of an Uncle Albert vibe.
(https://i.ibb.co/0pDRVNrY/greybeard.webp)
Quote from: David L on August 31, 2025, 09:02:33 AM"Remove the scratch plate" ;)
(https://i.ibb.co/hR5H3cy4/notreble.webp)
This is Gemini's effort at "remove the scratchplate". It's also removed the bridge pickup, as you can see.
Impressively, it's also removed one set of volume and tone controls, which is exactly what you'd have on a single pickup guitar.
Funnily enough - this guitar would suit me down to the ground.
It's absolutely seamless - you can't see a transition between the synthetic part of the image and the source. If you photoshopped it you'd never get the strings quite right.
By contrast: I tried exactly the same thing at https://huggingface.co/spaces/multimodalart/Qwen-Image-Edit-Fast and it left the source image unchanged.
Tried "please remove the white plastic part on the right side of the guitar body", same thing. No change. Odd.
Perhaps you should have asked - "Lerxstify this guitar" ? I'm sure that would've produced the desired result :-\ ;D
Yes just playing, but this is a field a few km from chez moi:
(https://i.ibb.co/21BtT3jV/field.webp)
"Please add a Space Shuttle on a launch platform, pointing upward, in the medium distance"
(https://i.ibb.co/TBHyLMP6/fieldshuttle.webp)
If you could only launch something like that from a few pallets...I'd want one in my back yard then
I asked ChatGPT the "what would happen" question .. it's not pretty.
T-10 to T-0: no pad systems- There's no lightning protection, no million-gallon-per-minute water deluge for sound suppression, no flame trench, and probably no proper hold-downs.
- Acoustic loads at liftoff without water suppression go well over 180 dB near the stack—enough to shred nearby kit, crack structures and rattle the orbiter's airframe and tiles to pieces.
T-6.6 s: SSMEs start- The three Shuttle main engines light first. Their exhaust impinges straight onto soil.
- Result: an instant steam/soil cannon—it scours a crater, throws rocks and mud back up into the engines and the orbiter's belly, and starts grass fires downwind.
- Without a trench, hot gas recirculates under the vehicle, roasting wiring and plumbing. Windows in houses kilometres away pop; the wind turbine's nacelle vibrates itself stupid.
T-0: SRBs ignite- Two solids come up to full thrust in a few tens of milliseconds. There's no "abort": once lit, they burn for ~2 minutes.
- If the stack is on anything other than a proper Mobile Launch Platform with hold-downs, it rips itself free or topples. Even on a platform, asymmetric blast into a dirt crater can kick the base sideways.
Most likely branch A — RUD on the ground (seconds)- SRB exhaust (plus soil, stones) sandblasts the boosters; a case gouge or joint leak becomes a catastrophic burn-through.
- The external tank ruptures in the thermal/structural chaos. Liquid oxygen/hydrogen mix and detonate; the solids fragment into several flying, still-burning chunks.
- Fireball height: hundreds of metres. Everything within a few hundred metres is ablaze; trees and hedges go in one go.
Less likely branch B — it clears the "pad"- It claws into the air trailing an enormous brown-white plume.
- The acoustic and overpressure still flatten outbuildings, smash windows for kilometres, and knock the turbine. Livestock panic; eardrums for anything close are done.
- Range safety doesn't exist here; with populated areas around, you'd have to self-destruct rather than let a 2,000-tonne stack with uncontrollable solids fly an off-nominal trajectory. The destruct turns each SRB into a giant, shredding Roman candle; debris and unburnt propellant rain over many square kilometres.
- If—by miracle—it flies a nominal 2 minutes, the spent boosters separate and fall on land, each the size of a lighthouse, with impact craters and secondary fires from residual propellant.
Aftermath (minutes to hours)- The SRBs produce hydrogen chloride (HCl) in vast quantities. You get a sharp, acidic cloud drifting downwind: metal corrodes, paint peels, vegetation browns, and ponds turn briefly acidic. Wildlife takes the hit first.
- Emergency services can't get close for a while due to heat, debris, and toxic plume. Power lines are down; minor roads are impassable with grit and shrapnel.
Days to months- Environmental cleanup: contaminated soil removal, dead fish/birds/livestock disposal, ash and alumina dust everywhere. Insurance nightmare.
- Investigators write a report whose first line is essentially: "Don't try to launch a Shuttle from a field."
Hmmm. Maybe not then
I've just noticed that it redid the wind turbine! How strange.
Just done a bit of reading about this and although the background looks identical, or very close - the whole thing has been re-rendered. It's a completely synthetic image.
"Certainly! Here's your Lerxstified guitar".
Fascinating. It's applied a headstock (cropped out of the original image) with the nut at the third fret. Looks a bit more like a ukelele. And the inlays at the top are now stars.
(https://i.ibb.co/M5s111cC/lerxstified.webp)
Oh my god that is horrid. What's next, a whammy bar?
This all makes me think - when someone posts a pic of their new guitar on a guitar forum, well - how can anyone trust that it's really what it purports to be? Would be a simple matter to ask an AI to transform a £350 Epiphone into a Custom Shop 1959 Les Paul reissue. Actually I think I'll try that.
AI image / video generation has changed everything really and there are of course much more sinister applications.
Quote from: Slim on September 03, 2025, 06:52:11 PMThis all makes me think - when someone posts a pic of their new guitar on a guitar forum, well - how can anyone trust that it's really what it purports to be? Would be a simple matter to ask an AI to transform a £350 Epiphone into a Custom Shop 1959 Les Paul reissue. Actually I think I'll try that.
AI image / video generation has changed everything really and there are of course much more sinister applications.
I think real guitar geeks (I'm not one) would pick out things that were incorrect.
Quote from: Slim on September 03, 2025, 04:41:36 PMI've just noticed that it redid the wind turbine! How strange.
"Vibrates itself stupid" is quite a human thing to say. Not the choice I'd expect from the artificial.
Was watching a YouTube video of a Bowie gig, or part of it. Indifferent video quality but I wondered if ChatGPT or Gemini could synthesize a decent quality still image from it.
Passed ChatGPT this screen capture:
(https://i.ibb.co/5XZ0Jj7F/bowie-vid.webp)
Got this back:
(https://i.ibb.co/GhtND4V/ai-bowie.webp)
It's not perfect, but this was its first attempt. He's playing a 12-string in the original image, ChatGPT seems to have invented an 8-string for him, although what interests me most is that the perspective has changed; the whole guitar has come out from behind the music stand.
Whatever chord he was playing has changed
Yes, Bø7(♭9, no 7th). Brave choice.
Perhaps I want to reuse an old photo on a "what are you wearing today" thread on a watch forum on Friday. Ah .. but it has a datewheel.. and it's not the 5th on Friday.
(https://i.ibb.co/HTJrF2Q/231215cari.jpg)
"Please change the date shown on the wristwatch from 5 to 12"
(https://i.ibb.co/9HbJL6Ts/smithsmod01.webp)
Curiously it also changed the position of the second hand and moved my wrist slightly.. look at the pattern on the laminate floor near the one o'clock position on the watch.
Also it got the "SMITHS" logo basically right, but it garbled the "ENGLAND" underneath.
I used Gemini.
"Please change the background to the surface of the Moon. The wrist should be clothed in a spacesuit and a Lunar Module should be visible in the distance."
(https://i.ibb.co/5gtfgpHR/carib-moon.webp)
Tried that with my Hamilton watch just now. Change from Wednesday to Friday.
Of course Gemini first gave me directions to actually change the day myself. But then I specified it needed to change the image itself. It did not garble the other text, and it did not touch the hands.
(https://i.ibb.co/sprFXfsY/PXL-20250910-142649613-MP.jpg)
Changed it into this, which is spot on.
(https://i.ibb.co/Kp08KHKm/1757514658202.png)
Did move the second hand this time: "please zoom out a little and include the Eiffel Tower in the background". I used the first image as the source.
(https://i.ibb.co/nMVCPPxx/eiffelhamilton.webp)
I have been to Paris during my holiday! Fantastic.
it did make up a piece of arm though, the hand looks right, the arm does not. I guess it was made part of the background.
Quote from: David L on October 05, 2024, 11:07:16 PMAhem, you haven't got all of your bike in btw
;D ;)
Well, I revisited this post less than a year later because I thought that Gemini might make a decent job of imagining the rest of the bike.
The original image:
(https://bikediary.uk/images/241005naseby.jpg)
"Please uncrop the image so that the rear wheel is completely visible"
(https://i.ibb.co/R4zS2k0p/gem-uncrop.webp)
An additional gravestone as well?
Quote from: Nick on September 13, 2025, 08:03:35 PMAn additional gravestone as well?
Well, it's been a year...
Quote from: Nick on September 13, 2025, 08:03:35 PMAn additional gravestone as well?
Yes - quite impressive I think. At the actual scene there's a path roughly where that extra gravestone is (just looked on Street View, it's Church Street at Naseby).
Actually it feels a bit like sci-fi, I find it amazing. Imagine what it'll be like ten years from now.
Wonder if Ai can advise on what is inscribed on the headstone and what was their life story?
I found an old wedding photo from my parents. Late 60s, large format (a4) and I asked Gemini to colour it, true to the style and age. I was stunned.. truly magnificent. I could have asked to take out some background elements but decided it should stay as true to the original as possible.
(https://i.ibb.co/NfPmwLq/PXL-20250917-073255863-MP.jpg)
(https://i.ibb.co/spTKDBfD/1758097681506.png)
That's brilliant, could almost be Kodachrome. Captures the colour film personality perfectly.
I've been tempted recently by an ad on Facebook to provide a short, coloured animated version of old still photos. Sorry the website name escapes me but it looks very impressive.
If I got AI to colour my parent's wedding photos - from the 50s - they'd still come out grey. It's grim up North...
This is cropped from a school photo from September 1971.
(https://i.ibb.co/67WtrnS0/jg11.png)
I asked Gemini to colourise it and resize it. It's a bit uncanny. I specified the colours and it got them all right, except that my eyes were too light brown. So I got it to redo it. Then I added a little bit of noise in the GIMP because it looked unnaturally smooth. A bit of film grain. I'm just gobsmacked by this.
(https://i.ibb.co/3m1Mz7Cc/schoolcrop.webp)
It really is unbelievable what this can do. I recall watching old episodes of say, CSI, in hich a tech blows up a video still and - low and behold - they discovered an object until then unnoticed in razor sharp detail. Impossible 25 years ago, but now.. here we are.
(https://i.ibb.co/ZRy3yVHF/1766672798266.png)
This is a photo I took at an airport in Spain in 1973:
(https://i.ibb.co/RpL2m51r/airportbefor.avif)
I got ChatGPT to remove the reflections on the glass, the specks of dust and the obstruction:
(https://i.ibb.co/v4S123wn/airportai.avif)
Came out pretty well I think.
I sometimes knock up a few cycling stats charts at the end of the year in Google Sheets. This time I thought it might be quicker to get an AI to do it. I took a screengrab of the appropriate portion of my spreadsheet and asked ChatGPT for a doughnut chart.
(https://i.ibb.co/7xPt5qnr/distances.avif)
OK - exactly what I asked for, but a bit crude. I asked it to make it a bit more stylish.
(https://i.ibb.co/kg22D0Dj/chatgptdo.avif)
I was happy enough with this until I realised that it had hallucinated an extra 7.8% entry. It had another couple of goes but wasn't able to fix it. It explained:
(https://i.ibb.co/HThjRGSL/chatgptexpl.png)
I took the first image to Gemini, to see if it could do any better. Just asked it to make it more stylish, prettier. No hallucination this time.
(https://i.ibb.co/4bbvXB6/gem-distances.avif)
So - in the end it wasn't actually any quicker than doing it myself in Google Sheets, though it might have been if I'd gone straight to Gemini - and I had to check the result more carefully. But the technology is improving fast.
Did you specify which bike should have which colour?
Quote from: Thenop on December 31, 2025, 11:48:00 AMDid you specify which bike should have which colour?
No. I did think of that actually - green for the X, Red for the first Boardman and so on - but decided not to bother.
Quote from: Slim on December 31, 2025, 12:15:01 PMQuote from: Thenop on December 31, 2025, 11:48:00 AMDid you specify which bike should have which colour?
No. I did think of that actually - green for the X, Red for the first Boardman and so on - but decided not to bother.
So ChatGPT and Gemini 'decided' to use the exact same colour per bike. Curious, unless the engine is the same.
Quote from: Thenop on December 31, 2025, 01:29:40 PMQuote from: Slim on December 31, 2025, 12:15:01 PMQuote from: Thenop on December 31, 2025, 11:48:00 AMDid you specify which bike should have which colour?
No. I did think of that actually - green for the X, Red for the first Boardman and so on - but decided not to bother.
So ChatGPT and Gemini 'decided' to use the exact same colour per bike. Curious, unless the engine is the same.
That's because they were both working from the first ChatGPT image which already had the colours defined.
This pic of Neil at Advision in London came up on Instagram earlier.
(https://i.ibb.co/0R2PCVzP/neil1.avif)
It's pretty rough - some sort of print pattern and yellowed, perhaps due to the film being out of date or overheated. I asked ChatGPT to fix both issues. One of those cases where the AI is a bit more real than the original image. Judging by the datestamp you can just see on the original, taken three weeks after I first saw them in Newcastle.
(https://i.ibb.co/9m0PJnr9/neil2.avif)
Quote from: R6GYY on January 10, 2024, 01:04:11 PMI've had a few goes now just to experiment, but any text I use gets corrupted in the image, translated with extra or different letters. What gives?
Two years after this thread started things seem to be much improved on this point. When I was creating promos for Ravenstone Radio I was getting text like "RAVSSTONERRADI" but now generally the various image AIs get text right. I thought about this when I got an AI to generate this pic:
(https://i.ibb.co/RTnXJ4Jq/gptwatch.avif%5B/ing%5DI%20wanted%20to%20see%20what%20the%20wattch%20(a%20Casio%20A1100)%20would%20look%20like%20on%20a%20NATO%20strap.%20Not%20bad%20as%20you%20can%20see%20but%20crucially%20it%20got%20all%20the%20text%20spot%20on.Was%20curious%20about%20this,%20so%20I%20asked%20Gemini.%5Bimg%5Dhttps://i.ibb.co/XrmwCCqx/badtex.png)
Where was I? Oh yes. So I was curious so I asked Gemini why things had improved. New method?
(https://i.ibb.co/XrmwCCqx/badtex.png)
(https://i.ibb.co/YTNfgRvX/halfinvented.avif)
This pic of me was taken in 2010. But the bottom part of it has been imagined by the ChatGPT image model; this is the original:
(https://i.ibb.co/nqz7pjhK/half.avif)
(https://i.ibb.co/5hNkbSQL/old-pw.avif)
Just a thought for the Power Windows 40th anniversary edition, if there is one.
Quote from: Slim on February 06, 2026, 01:12:02 AM(https://i.ibb.co/5hNkbSQL/old-pw.avif)
Just a thought for the Power Windows 40th anniversary edition, if there is one.
Lifeson's not aged well😀
Been tinkering with image generation on my own PC. You do need a pretty good GPU for this, but I asked Stable Diffusion to generate an image with the prompt:
A photorealistic, cinematic portrait of a stunning, sophisticated woman in her early 40s with long, sun-kissed hair, wearing sunglasses, standing on a secluded Mediterranean beach. She is wearing an elegant, classic navy blue swimsuit. The scene is captured during the golden hour, with warm, low-angled sunlight highlighting the texture of her skin and the soft glow of the sand. The background features a gentle surf and distant coastal cliffs in soft focus. 8k resolution, professional photography, natural and graceful composition.Natural skin texture, subtle pores, slight film grain
There are lots of parameters and tweaks in the controls. I used a model called biglustydonutmixNSFW_v12.
I very slightly increased sharpness in the GIMP but it looked very good straight out of the AI.
One thing that the Stable Diffusion models don't do well is human eyes, and that's why she's wearing sunglasses.
(https://i.ibb.co/cndxNdb/beachwoman.avif)
Tried to create some art on my own PC. Here, I just asked for an impressionist masterpiece with cinematic lighting:
(https://i.ibb.co/8nHQrgyG/venice.avif)
.. and here, "London's West End by moonlight, empty streets, oil painting"
(https://i.ibb.co/1JqHFshn/image1.avif)
OpenAI's image model (via ChatGPT) has had an upgrade. Used the same prompt as I used on my own PC:
A photorealistic, cinematic portrait of a stunning, sophisticated woman in her early 40s with long, sun-kissed hair, wearing sunglasses, standing on a secluded Mediterranean beach. She is wearing an elegant, classic navy blue swimsuit. The scene is captured during the golden hour, with warm, low-angled sunlight highlighting the texture of her skin and the soft glow of the sand. The background features a gentle surf and distant coastal cliffs in soft focus. 8k resolution, professional photography, natural and graceful composition.Natural skin texture, subtle pores, slight film grain
(https://i.ibb.co/9HBKkJXG/openai2swimsuit.avif)
Quote from: Slim on February 06, 2026, 01:12:02 AM(https://i.ibb.co/5hNkbSQL/old-pw.avif)
Just a thought for the Power Windows 40th anniversary edition, if there is one.
I imagine the belly would be a little wider and the screens a little flatter ? 😃
Back in the '80s I had a paperback copy of Philip Dick's The Man in the High Castle, the one with this cover.
(https://i.ibb.co/PzFszn7F/highcastle-orig.jpg)
I just tried to describe it from memory to ChatGPT, to recreate it:
A book by Philip Dick called The Man in the High Castle is about an alternate reality. In the end it turns out that it is not real, it's somehow been superimposed on the real world. In this alternate reality the US has been defeated by the Japanese. Can you please generate a book cover idea - portrait mode - a man at night, in San Francisco. Modern buildings. He looks uneasy, unsettled. Out of place. Behind him on a high billboard is a Japanese drink advert with the text "The Spirit that Conquered the World" - Japanese themed, but in English.
It came up with this, which I think is brilliant.
(https://i.ibb.co/TMdCcTK1/gpthighcastle.avif)
Whisky from a can!
It's a real thing!
https://www.drinksimplicity.com/spirits/premium-bourbon-whiskey-355ml
(https://i.ibb.co/MyD8Qsz5/utahwhisky.jpg)
Tesco also sell Irish whiskey in small cans.
I stand atop a spiral stair
An oracle confronts me there
(https://i.ibb.co/KcFppTTh/oracle.avif)
Lovely image, nice interpretation of a spiral stair.
Amazing ai imagination.
This is Gemini's version. It took a few iterations to get it closer to the idea, hinting at the next two lines
He leads me on, light years away
Through astral nights, galactic days
Again the model chose a female oracle but I really like that so I opted to keep it. Actually I think Neil missed a trick there.
(https://i.ibb.co/7xVGB67n/atop.avif)
Grok's effort
(https://i.ibb.co/KxLtmfz4/grok-oracle.avif)
Quote from: Slim on February 06, 2026, 01:12:02 AM(https://i.ibb.co/5hNkbSQL/old-pw.avif)
Just a thought for the Power Windows 40th anniversary edition, if there is one.
How did I miss this? Fantastic! Would it be possible for us to all have our image on this? At last I forgive AI.
I don't know what gave me the idea for this while I was out on a bike earlier, but I asked ChatGPT for a single page of a late '60s Superman comic.
There was one slight error which I fixed - one of the speech balloons had two tails instead of one. Edited it out in the GIMP.
(https://i.ibb.co/QvPL0q2F/ai-superman.avif)
As part of my, er, 'hobby', a few months ago I was experimenting with using AI to upscale some old porn clips.
I thought why not use AI to create some DVD inserts and labels. Sadly, I no longer have the prompt which is annoying.
Still, here are the DVD insert and DVD label. If I did it again, I would try and do it so that the chap in the pic keeps his face intact on the DVD label. All in all though I was quite pleased.
(https://i.ibb.co/k2327s00/Volume-1.png)
(https://i.ibb.co/V01qyLJB/LAbel.png)