Forum Discussion
Week 1 Discussion
Which AI Assistant feature felt most intuitive or useful to you during the practice activities?
Using the AI text-to-speech feature. It is very easy to use, whether it is to generate or revise the audio. One thing I have noticed though is that on occasion, when generating a lot of audio via Storyline and/or keeping the app open across desktop sessions, the AI dialog box disappears behind the main window. I have to close the application and reopen it to eliminate the behavior.
What challenges did you face when trying to get AI to produce the results you wanted?
I have found that the AI image generator will on occasion show the same individual multiple times in the image or will often show multiple and sometimes all the individuals in the image in blue button down shirts or blue suits if you specify an office or corporate setting.
I have had that issue too where the image showed the individual multiple times or some strange language. I closed it out and redid it and it didn't show the same individual the second time.
I too have had that issue. Also, when I do want the same individual but with different facial expressions, I can't seem to get good results.
Yes, it would be nice if we can take an image that was generated and then iterate on it with the AI assistant. Sometimes I like an image, but there's just one thing I'd want to tweak about it so that seems to be a limitation with the tool right now.
I agree and really love the AI Text-to-Speech feature. I also love being able to change the voice style and version. I've found the best results with V3 and it sounds very realistic when speaking.
I agree, the AI text-to-speech feature is very intuitive. It’s quick to generate and even quicker to revise, which makes it easy to update narration without re-recording everything.
I’ve noticed something similar with the image generator, which tracks with every other image generator- sometimes you get “same-face syndrome” where multiple individuals in a crowd look nearly identical. I’ve also seen that tendency toward uniform clothing (like everyone in blue button-downs in a corporate setting). It feels like a form of mode collapse where the AI defaults to a narrow interpretation of the prompt. When that happens, I’ve found being extremely specific in the prompt or manually adjusting in Photoshop afterwards is usually faster than regenerating multiple versions.
I totally had this when creating a course on workplace harassment. In a conference room picture there were 8 people, 4 were the same man with slightly different hair styles and one woman with a full, nicely cared for blond beard.
Agreed, it seems like the more detail you give this image generator the more likely it is to latch on to the wrong part of the description... I have gotten some very unusable images with too much detail in this tool. (Any other tips from superusers on how to get better images from this tool?)
I've noticed this too. The same bearded dude sitting across from his twin! I'd be interested to know the best way to prompt the Ai image generator to avoid this.
I agree that the audio generator is one of the strongest features of the AI assistant. I think it's great that Articulate was able to partner with Eleven Labs to utilize their AI voice technology. I love the variety of voices and that ability to customize the audio options.
Hi there! I noticed the same thing when reading your response to the second question, and I completely agree with you. It would be helpful to see more variation across the four images that are generated so there is a wider range of options to choose from. Since the results are driven by the prompt, it seems like refining the prompt could help improve that variation. I also wonder what might happen if I asked for multiple different shirt options or other small visual differences to see if that encourages more variety in the results.
