That’s interesting.
I don’t have an Adobe subscription either
I got bitten hard by the AI imagining bug about three months ago. To the extent that I all but stopped doing any work in Rhino7 and even much longer since I had done any rendering work in Rhino. I was spending so much time beating my head against the steep learning curve that I even stopped checking in here on the forum for Rhino. The majority on my Rhino work is doing marine and yacht design and I have worked hard for 10 years on my rendering skills especially. It takes a lot of work to create enough detailing as mentioned above to make convincing photo-realistic renderings of watercraft and the water interface and environment is an especially challenging task. My thought when I started exploring AI imaging was to help and apply some of the mindblowing techniques from AI to my Rhino model 3d rendering. Especially with interiors as lighting and textures are so much more important to the look desired. I was able to get some nice AI results but found that I had little control just where design elements would show up like galley faucets sprouting out of bulkheads and the chef standing in the ice chest.
I didn't have much better luck with the rigging details for sail boats and without using a 3d model and a render there to use as a reference image it got pretty frustrating. I tried using the well known and prolific yacht designer Bob Perry (my main client) into my prompts in this manner 'in the style of Bob Perry' much like one does to emulate an artists style and work but apparently boat and yacht designers names were not used to 'train' the AI databases. So it may be a while until I can really use AI in my Rhino3D yacht design efforts but at the rate the AI imaging world is evolving I am going to try and keep up with the pace. I got a notification while still in bed reading my news feeds of the release of MidJourney V6 last week and had it up and running just as soon I could put on a pot of coffee and join the upgrade beta or alpha of the day. Now there is another incremental update maybe tomorrow and one promised each week until the end of January! I still feel like I am paddling up the back side of a surfing wave having only gotten into the game three months ago but hopefully can get on the frontside of the coming 'set wave' and not get left behind.
It was great to get the 'synopsis' from McNeel today with Thomas An's thread on the subject and was very encouraged to see the response especially from Kyle and Sky who always seem to be open to whats new in Rhino. Let's keep the community effort going here in Rhino as that openAI co-operation and sharing is what is behind the astounding progress in AI imaging.
Here is an early AI attempt for a big condomaran design. Seems like big top heavy overladen designs are getting all the attention in the real world which may explain why the AI engines can summon up something to start their magic with.
Here is an attempt at an interior wide angle shot using AI that has a nice look but I had little control of the actual layout.
Here is a result at trying to show a hostess serving French Toast (long story) on the alfresco dining deck of a large yacht. I think the prompt ‘large yacht’ was responsible for the very large French Toast. Lots of quirks to sort out and may get frustrated for us Rhino control freaks, but the amazing image quality for human crew figures is a big plus for AI.
Look at this Mid-Journey 6 result for a Tahitian Vahine that I am very pleased with.
1 Like
Just to show the possibilites of using AI imaging for my yacht design rendering here is a mega yacht with a prompted south seas environment that I did early in my AI addiction. I guess mega yachts got better represented when the AI database training took place and a fully Rhino3d render from the 3D model would be a big challenge compared to this. But the multiple rainbows are a giveaway that this is an AI ‘pretty picture’ as opposed to a physically correct ‘virtual model’. I am having a heck of a time getting AI to understand that oars on a Whitehall skiff come in pairs. I only get one oar for my subject and if I ask for ‘two oars’ I get tow crew with one oar each. Maybe that would work for a rowing shell though!
My AI skills must be improving in the last few weeks as I just had another attempt at re-creating the big yacht posted above in a different AI rendering engine to get this. I included the UI to share the prompt used and it took me less than a minute to cook all of this up including the time I spent thinking of how to phrase and write the prompt! Less time that it took to write and post this…
Imagine trying to create that wake in Photoshop, much less in Rhino.
Thanks for introducing this thread and re-kindling my efforts at applying AI into my Rhino3D yacht design and modeling workflow, lots of good comments and thoughts from the gang here. I’m going to pay up for another month of MidJourney 6 server time tomorrow and have another go there and dig up an old Rhino model to do a side by side comparison. These last few I have dialed up tonight are using Mage.space which I have the most time with so far. I haven’t spent anything on Mage just using their free account but it serves well to work out the bugs in a prompt before using in a paid account such as MidJourney. Either way once a prompt is dialed in it is so quick to do further ‘variations’ or iterations which is where one can really let the client get to play ‘what if’ which is where one can really earn some money an sell the client.
One more before I turn in...
Well, maybe two. Like I said, once you start getting results it get addicting! Now I will probably be up half the night…
2 Likes
I’m in awe of the render projects shown here. Are these all for actual product previews? Or also just for 2D ads and such? I feel dum for not exploring advanced rendering yet, but feel grateful that for jewelry, thus far the RhinoGold6 Render Studio has sufficed, and the Rhino Raytrace is pretty good too. 2024 will be a good year for learning some new skills.
Working on getting an old Rhino3D model yacht for comparison renders with the AI possibilites as mentioned yesterday. I did find that AI seems to like the prompt ‘mega-yacht’ even more than ‘large yacht’. I guess when the trainers started ‘scraping’ the web for images they started at the top…
This is a mere ‘super-yacht’
The first one looks a little nicer to my eyes.
Even though the AI does a lot of the work, the composition of the prompt and the taste of the artist (by selecting / filtering the images) place some “soul” back in the machine and play a large role in the outcome.
I took a closer look at your comparison of the orange juice renders and had a go at it myself. I did these in Distillery which is the entry level free generation tool on Discourse. Not to bad for starters but has a sort of flat plasticy look and feel that I think I can improve on when I pay up for credits on Mid-Journey.
1 Like
There are people who use it to make money today. From T-shirt designs, logos, Tattoo ideas, fantasy works. Anything organic and fluid works well. Someone even created an AI female instagram model and makes $4k a month through only fans.
… but for industrial design work (as mentioned at the beginning of the thread) you can’t guide the details with precision. As such, it is best for design inspiration and lighting moods, so you can try to get your own models to look better.
For example I was playing around with Jewelry prompts and it started giving me some ideas I hadn’t thought before:
Can be used to get ideas for fashion design. When I did this image a lot of people ignored the Moai, but loved the costume.
Was designing a logo and used the AI to rough out the proportions and details from this as a start:
3 Likes
You changed the scene and lighting though. The whole point, originally, was to feed it something I had already done and see if it could give me ideas to improve the overall composition.
I signed up for a model sharing thing years ago called Modelo. It promised cross platform interconnectivey and the usual ‘buzz word’ stuff. I uploaded a couple of my sailboat models in the hopes I could get Bob Perry more involved in using a 3d model in our collaborations but the program didn’t do much for Bob, or me for that matter. I haven’t touched Modelo for at least 6 or 7 years but I must have still been on their mailing list. I got an announcement a couple of days ago proclaiming that they had AI rendering from SketchFab, Rhino3d, obj and the other 3d formats so I figured out that I would have a look. I uploaded a recent project of a hard dodger top for an existing Perry design that I had the owner do an IPhone lidar scan of the ‘as built’ deck and cockpit since the original 2d drawings were on paper and the scans from Bob were not worth trying to reverse engineer a 3d model in Rhino. The scans worked out surprisingly well and I had the project sitting open on my desktop since the owner was building a 1" foam panel ‘mock up’ before building the real thing. I uploaded the scan and my simple Rhino form to Modelo and tried the AI render feature and it interpreted the dodge and deck as an auto or golf cart and the AI generation was hilarious. Not even worth posting. But I had another old Perry design in full 3d for the hull, deck, house and cockpit so gave Modelo a second chance. Barely worth sharing here. Don’t hold you breath waiting on Modelo to learn what a sailboat is.
I’m staking my hopes on MidJourney and narrowing my curiousity down to just a few software solutions as there are too many pretenders in the AI image generation right now. Everybody and his brother are trying to jump on the boat so to speak!
I have the Rhino model of this project open and am trying to use screenshots from the Rhino shaded mode windows and use as a reference in AI with little success, so far.
Are all these generated by AI?
Which ones are made/photographed by human?
If you mean at the first post, in each pair one is a man-made render and one is AI.
So the left one is man-made and the right one is AI- generated?
No, each pair is random order.
And it is intentional that you don’t disclose which one is made by AI?
I was goofing around with another (of many) AI rendering engine today and was creating a Rum Bottle image and the powers that be in the AI decided to thrown in an orange juice version of a Dark and Stormy drink! Is it real?
1 Like
I don’t drink, but I guess OJ+Rum is a thing: 3 Rum and Orange Juice Cocktails That Make Brunch Better
I’m surprised to see this thread slowing down, must be the rum sundowner cocktails! I finally was able to generate an image of a small boat design that I would consider accurate enough to rival my Rhino3D renders. One of my design clients/mentors has been telling me to forget about using AI on our joint projects, that AI leaves him cold. He is urging me to stick with our long established workflow of my using his 2D CAD to model Rhino3D ‘virtual models’ and then render from the actual NURBS or render meshes. But he doesn’t realise just how much work and creativity it is to set up render meshes and mapping and realistic environments especially for boats in convincing water. I am going to share this image with him and see if it leaves him ‘cold’ or if he even realises it is an AI image!
2 Likes
That does indeed look pretty convincing. Let us know what your client says 
I am a little hesitant to show my client, as cold as it has been here I don’t want to leave him ‘colder’. I have been using what seems to be the most capable ‘top shelf’ AI engine MidJourney and wanted to compare with the previous ‘girl in a peapod’ image I posted here. This looks looks to have better detailing of the boat itself and I still have to get the girl to turn around and sit at the oars but I guess the prompt term ‘double ender’ is out of the AI’s realm of understanding… Thanks for the compliment Martha. Are you part of the famous Dijkstra boat building dynasty?
1 Like