Vectorworks, Inc Employee Popular Post Dave Donley Posted March 20 Vectorworks, Inc Employee Popular Post Share Posted March 20 Hello forum! I would like to hear from you about the 2025 Update 4 AI Viz changes. This version adds 5 major new things - style transfer, aspect ratio control, inpainting, improve details, and direct creation of Vectorworks image fills, backgrounds, textures, and image props. I hope these new additions make the feature (much) more useful to you. Please let us know how it goes, thanks! 5 Quote Link to comment
Vectorworks, Inc Employee Dave Donley Posted March 21 Author Vectorworks, Inc Employee Share Posted March 21 Here is a video describing the latest features and how to use them (I trimmed the demo video we use internally for new features). Hope this helps you get started. Vectorworks 2025 Update AI Visualizer Features.mp4 2 Quote Link to comment
Flair-Studio Posted March 24 Share Posted March 24 Hi Dave, I tried to play with the new AI vis, and although some features like the draw mask are very cool additions, the overall experience is still not so good. It seems slower than the previous, it doesn't refresh the base image the way I would expect to, and often when I ask to improve the details strength it just gets stuck. The video you have posted is useful but I hope there will be a cofee break soon! Thank you Federico Quote Link to comment
Vectorworks, Inc Employee Dave Donley Posted March 24 Author Vectorworks, Inc Employee Share Posted March 24 Hi again @Flair-Studio can you elaborate on this so we can make it better: "it doesn't refresh the base image the way I would expect to" , is this composition image or the inpainting features, and how is it missing the mark? The timing between Update 3 and 4 is within a second of each other, but in Update 4 the user experience is a lot more "sticky", so generation times are important since you experience them a lot more in Update 4. Quote Link to comment
Flair-Studio Posted March 25 Share Posted March 25 Hi Dave, Thanks for coming back! My process is quite basic, when I need to explore options with the AI, I usually copy paste the base images in a new file, then I select them one by one and open the AI with the image selected. With the new interface, under composition, I need to go to refresh in order to use one image which is already in VW. Well I had experienced some issues in refreshing the image. The introduction of the edit portion of the new AI is a great and long expected one, finally once there is a lucky strike with the generation portion it is possible to refine the outcome a bit. It would be great to be able to undo part of the painting without having to erase the hole mask, like a negative brush... I also wanted to ask, in the edit section of the new AI, what is the difference between the Add and the Modify buttons? It seems like they are doing the same funcion once you have painted a portion of the image? Thanks again and don't get me wrong, the new AI is definitely a great improvement, and it would be also great to see a new coffee break on the new AI coming up soon Quote Link to comment
Vectorworks, Inc Employee Dave Donley Posted March 25 Author Vectorworks, Inc Employee Share Posted March 25 Thanks for the additional info! The refresh will use selected bitmap object or selected rendered viewport, or will make a screenshot of the whole Vectorworks window if a bitmap or viewport aren't selected. The Remove button removes areas that you have masked, is that different than the negative brush you mention? The add does a strong add, replacing what is under the mask. Modify considers what is in the masked area and depending on strength value will modify it strongly or weekly. A weak modify would tweak a few details without changing the basic region. A strong modify approaches what an Add does. Modify can be used for example to modify the color of a window without changing the frame shape. HTH 1 Quote Link to comment
Flair-Studio Posted March 26 Share Posted March 26 Negative brush is a nice to have, to have the ability to refine a complex mask Sometimes it takes forever to use the improving details section: am I doing anything wrong? Quote Link to comment
Vectorworks, Inc Employee Dave Donley Posted March 27 Author Vectorworks, Inc Employee Share Posted March 27 @Flair-Studio Improve details takes a few moments to get going, shouldn't take much more than a minute total. Quote Link to comment
Pat Stanford Posted March 28 Share Posted March 28 I have been playing with this today and have a question and an observation. The observation: When you Refresh the starting image, it does not use the selection, but rather everything in the drawing window. So if you are zoomed out, you will get a lot of border fill around whatever you are trying to visualize. So you want to zoom in, and possibly change your drawing window size to get the image you want to feed to the visualizer. The question: How accurately does the mask need to be drawn for Add/Edit? What happens if you can't draw a perfectly straight edge to match something you are trying to recolor? What about missing area in the middle of the mask? It is hard to see the mask against some backgrounds. Overall fun and useful for what I am doing right now. Quote Link to comment
The Hamma Posted March 28 Share Posted March 28 Everyone: Please post your favorite prompts and settings. 1 Quote Link to comment
Vectorworks, Inc Employee Dave Donley Posted March 28 Author Vectorworks, Inc Employee Share Posted March 28 23 minutes ago, Pat Stanford said: I have been playing with this today and have a question and an observation. The observation: When you Refresh the starting image, it does not use the selection, but rather everything in the drawing window. So if you are zoomed out, you will get a lot of border fill around whatever you are trying to visualize. So you want to zoom in, and possibly change your drawing window size to get the image you want to feed to the visualizer. The question: How accurately does the mask need to be drawn for Add/Edit? What happens if you can't draw a perfectly straight edge to match something you are trying to recolor? What about missing area in the middle of the mask? It is hard to see the mask against some backgrounds. Overall fun and useful for what I am doing right now. Hi @Pat Stanford If you have a bitmap object or rendered viewport selected it will use that directly. Or you can import a file from disk. Otherwise we take a screenshot of the current window. It does not have to be super accurate unless you need it to exactly follow an edge. You might fill in interior with large radius brush then use smaller brush to get to corners. It is a top wish to do lasso and/or polyline mode for masking we will add that later. Quote Link to comment
Pat Stanford Posted March 28 Share Posted March 28 OK. That makes sense with what I was seeing. I had a bitmap on top of which I drew a couple of polys. If I had the bitmap selected, it only got that and ignored the polys. With the bitmap deselected it got everything. Eventually I figured out to shrink the drawing window to "crop" to what I wanted. Thanks for the update on the mask. Could you please also add to the wish list to add constrain horizontal/vertical with the Shift key also? 3 Quote Link to comment
Vectorworks, Inc Employee Dave Donley Posted March 30 Author Vectorworks, Inc Employee Share Posted March 30 Yes will do @Pat Stanford would a rectangular marquee serve? Quote Link to comment
Pat Stanford Posted March 30 Share Posted March 30 Marqee would help, but I think vertical/horizontal constraints are needed also. If I need recolor the tubes in the front or the control panel in the back I need to be very careful to try and get the vertical edge. A marquee is not especially useful for objects in perspective like this. A polygon marquee where I could define the edges and have the center filled in would be more useful. 1 Quote Link to comment
Flair-Studio Posted April 5 Share Posted April 5 @Dave Donley: I saw that on the other AI chat, the most common one, it is possible to reference multiple images in the prompt engineering process, so that the outcome is going to get much closer to what you envision. In example, for an interior rendering, in the other AI you can provide the image of your model but then you can add a number of images referring to each furniture, light, or artwork element within your rendering. This gives actually much better result and doesn't feel like rolling the dice anymore. Is this something in development for VW AI as well? In other words, what do you suggest to do in terms of prompt engineering with the new VW AI to get the results that you want, other than reducing creativity,etc? Thank you Quote Link to comment
Vectorworks, Inc Employee Dave Donley Posted April 7 Author Vectorworks, Inc Employee Share Posted April 7 Thanks @Flair-Studio which "other" are you talking about? Quote Link to comment
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.