Tired of Code Errors from Your ChatGPT Blender Add-on? Here’s the Fix!
Joe Morris introduces his new Ultimate Script Generator for Blender 4.3.
https://youtu.be/b2EJN_OAcUY
Our cutting-edge script generator harnesses the power of OpenAI’s advanced technology, using the exact same model available with the ChatGPT Pro subscription on the ChatGPT website, to create scripts from your detailed user commands. If a generated script encounters an error, our system automatically tests it and, if necessary, sends it back to the OpenAI model for revision. This iterative process, which defaults to five iterations, works in a loop: it reads your user command, generates a script, tests it for errors, looks up the Blender API, manual, Blender Stack Exchange, and more, and then sends it back to the OpenAI model for revision.
Notable features include:
- OpenAI-powered script generation for automation.
- Automatic error testing and revision.
- Boost User Command feature upgrades your command to be Blender Manual-compatible.
- Verified compatibility with Blender 4.3 only (but will convert scripts to Blender 4.3!)
- Automatically locates example scripts from our database based on errors.
- Easily updates older scripts to the latest Blender version.
- One-click error fixing in the text editor.
- Allows you to provide your own code examples.
- Can find/pull (and fix as necc.) code examples from Stack Exchange.
- Optimized functionality for creating animation scripts.
- Optimized functionality for creating small/medium shader/compositor/geo-node setups.
Deep Dive:
https://youtu.be/_CaFP8KjZz0
And one final caveat that’s also a bonus: in some cases, our system may encounter an error it can’t automatically resolve. When this happens, we’ve made it so simple that anyone can resolve these errors themselves, even non-programmers, using our easy-to-follow instruction set system that’s so straightforward even kids* can create working scripts for Blender.
This is a major bonus as programming has been around for decades and nowhere else, as to our research, can non programmers guarantee error free code!
*(Kids 6th grade and higher as there is some required reading of techy stuff.)
The website is giving me seizures LOL!
And I’m not sure on all this AI stuff making it’s way into Blender……
I totally understand what you mean—there’s a lot of AI stuff out there these days, but the great thing is that most of it works the same: you type in a command, and it just does the thing you need. That’s exactly what out system brings to Blender. You type in anything you want, and it generates a script and fixes the errors, to make that happen
Thank you for replying!…If you’re referring to our website, btw, please let me know what the issue is so I can address it! I want to make sure everything works smoothly for you.
As for AI functionality, it’s really exciting how far it’s come. Nowadays, it works just like it was imagined in the movies of the ’80s—you simply enter a description of what you want, and the AI does exactly that. We’ve brought that kind of functionality into Blender scripting. With our system, you can describe what you want in detail, and it generates scripts to make it happen. Please, let me know if you’d like to learn more or have any questions! We are always available to help!
Yeah… No.
thats bot what AI is doing.
it is not giving me exactly(!!!) What I type in.
If this would be the case, it would be great.
Go to an AI image generator, and you will get something… A vague approximation of what is typed in. This is impressive, no question.
But AI is all but an exact tool. If it would be, why using blender at all?
So with your coding automation it is the same, no matter what you promise. It is for sure not capable of solving all tasks.
As more and more complex the task is, the more it will fail.
Great, it can generate a cube… But do we need an AI for that?
That’s the thing is you must not have used ChatGPT much… it will do exactly what you say. The thing is, you want to give it extremely detailed commands. Now, if you try ChatGPT using the free version, you’re exactly right, you’re not gonna get a good result, but if you use the Pro version with their advanced model, it will do verbatim everything you say if you provide detailed commands. See, this functionality that I created has two different ways to use it.
The manual way is to give it an extremely detailed command telling it exactly what to do, like if you tell it to add a cube to this scene and then go into edit mode and then extrude a face of that cube in the Z direction .5 meters and then inset that face one-tenth the size of the original face, and then go back into object mode, and then add a viewport coloring, and then go back into edit mode, etcetera. It’s going to do all those tasks because we’re using the advanced AI model that’s in the ChatGPT Pro version.
We also have a boost prompt feature where you could just tell it, you know, roundabout basically, you know what you want without really knowing how to do the task in Blender, and it’ll look up the manual and convert your command into a list of manual-verified instructions. Then you just verify the instructions in the console, and if they do describe exactly what you want, then great, you can generate the script. But if they don’t, you could reorder the commands or add new commands. If you’re unsure how to do a specific thing and you don’t know what to tell it to get it to do that specific thing, you can even take a tutorial and find out how to do that. Then you know how to explain that in the future.
But keep in mind, the AI is very awesome at filling in the cracks and understanding what you mean. I’d be happy to provide a demonstration if you’re interested, like a demonstration video of the kind of detailed commands that I give. And it’s also really easy to use when you understand that GPT will understand even if you make voice typing errors. So you could just sit there with voice typing on and speak for 2 or 3 minutes, be very precise, tell it what you want, tell it what you don’t want. You can even give it example code, tell it to look up the Internet and, you know, find you supplementary information. And that’s on ChatGPT, The Internet look up that is,. within the add-on, we give it the ability to look up the Blender API and manual, but it automatically does that every time it gets an error. It checks the manual, checks the API, makes sure everything is done right… We haven’t added generalized Internet lookup capability yet but we’re perfectly willing to and will do that as soon as we get a feature request for it we’re trying to make sure right now the basic functionality is very very solid so people can count on it…
And again if it would help sell you on the process I’m happy to create a specialized video For you showing how to get perfect results from chat GPT see I’ve been using ChatGPT since it came out and have gotten very very good at it and all my knowledge I’ve poured into this functionality that I’m offering in this blender add-on so that people that aren’t first in how chat GPT works can still get awesome results by following the simple instructions that we provide in the add-on…. And in closing please don’t use the image output that you get on open AI as an exact representation of what it’s capable of because I haven’t really got the most perfect results out of that either but the code generation is top notch… There are other websites and YouTube videos that detail much more quality image generation and even video generation that generates image and videos perfectly there are people out there that have no problem generating images exactly how they want them to be..
And one more item in closing when you’re voice typing a command for chat GPT if it’s not extremely long you’re not gonna get the the best results…. CAI benefits from having context and detailed data sets on what you’re doing when you give it a long command you’re supplying it at data set giving it context and training it on your exact intention so even if you provide supplementary data to it maybe that may seem unrelated you know it could help give it context so whatever comes to your mind while you’re explaining what you wanted to do just throw it in you know it’ll give it more context about your current situation what you want the script for what you intend for it to do etcetera……you want a very long and very detailed command and again never worry about voice typing errors that’s what stops a lot of people from making those long commands is like thinking holy crud I gotta go ahead and now edit all that text now it can glean the context very easily
Just one more quick note I noticed how long my message was when I pressed the send button and it displayed on the web page and I do apologize for the length it might be a little bit of a difficult read but keep in mind the reason why I made such a long messages I get so used to talking to ChatGPT and giving it long sense of instructions that it just kind of spilled out here but that comment I sent represents the exact lengths of a good request to generate a script.
you lost me at “sends it to open ai”
Here’s the corrected and updated version:
—
Reply:
Sending a particular piece of data to the OpenAI model is synonymous with pasting that data into ChatGPT and pressing enter. For example, our system might send a request like this:
1. Create a cube.
2. Go into edit mode and extrude a face.
3. Move that extruded face in the Z direction by 0.5 meters.
4. Inset that face.
Usually, shorter scripts like this have a better chance of success on ChatGPT, so you may not get an error with this example. However, to understand what we mean, try beefing up the commands. Add steps like numbers 5, 6, 7, and so on, with additional tasks you’d like it to perform. Once you’ve created a longer set of instructions, run the generated script in Blender’s text editor.
If it gets an error, copy and paste the error message back into ChatGPT and let it know what happened. ChatGPT will then refine the script and give you a new version to test. This process can help you understand how the system works and the effort involved—and it’ll give you a better appreciation for what our product automates.
Our system makes it possible to generate scripts up to 350 lines long—up to 350 Blender commands!. It automates the entire process: generating scripts, testing them, handling errors, and refining results. This saves you significant time and makes scripting accessible to everyone, whether you’re a beginner or an experienced programmer.
Let me know if you have any questions or want to explore how this tool can transform your Blender workflow!
Sending a particular piece of data to the OpenAI model is synonymous with pasting that data into ChatGPT and pressing enter. For example, our system might send a request like this:
Create a cube.
Go into edit mode and extrude a face.
Move that extruded face in the Z direction by 0.5 meters.
Inset that face.
Usually, shorter scripts like this have a better chance of success on ChatGPT, so you may not get an error with this example. However, to understand what we mean, try beefing up the commands. Add steps like numbers 5, 6, 7, and so on, with additional tasks you’d like it to perform. Once you’ve created a longer set of instructions, run the generated script in Blender’s text editor.
If it gets an error, copy and paste the error message back into ChatGPT and let it know what happened. ChatGPT will then refine the script and give you a new version to test. This process can help you understand how the system works and the effort involved—and it’ll give you a better appreciation for what our product automates.
Our system makes it possible to generate scripts up to 350 lines long—that’s up to 350 Blender commands. It automates the entire process: generating scripts, testing them, handling errors, and refining results. This saves you significant time and makes scripting accessible to everyone, whether you’re a beginner or an experienced programmer.
Let me know if you have any questions or want to explore how this tool can transform your scripting process!
You still lost me at sending it to open ai…
Pasting it to chatgpt IS sending it to open AI
Yes I do apologize I should have been more clear in my original post here but chat GPT is a website created by the open AI company…. And they write a programming library that programmers can use to work with the artificial intelligence that they created and that library is basically just called ‘openAI’ so I get used to calling it that but when I’m making advertisements for my product I try and make it simpler for the people reading my advertisements as they usually would know it as chat GPT… So in a nutshell when I say sending it to the openAI model that’s equivalent to adding text to the box on chat GPT.com and pressing enter.. And also for some context the functionality that I created uses the same artificial intelligence model as you get on the chat GPT Pro subscription which is what makes this so awesome… And I forgot to say but thank you for commenting!
(sorry there are 2 messages for each reply, I was entering these replies on Ubuntu yesterday and wasn’t getting any indication that the messages were sent… I sent a message to the Blender Nation team to see if they would delete the extras for me.)