So there are models for Pose estimation like [PoseNet] or Qualcomm's [MediaPipe] that give you a graph of the body position. This metadata can be converted to a string representation and be fed into a generative model. If I record myself in different exercises, extract the poses, and label them with the appropriate steps, I might be able to perform few-shot inference or RAG over a set of labeled positions, such as the model can infer new body positions from exercise steps.
If that'd possible, then it is fairly easy:tm: to generate illustrations (e.g. like stickman figures. See [Kawamoto] for an example)
Of course, none of this is necessary if I can record (or let users record) and upload videos of each exercise, having enough of them to not feel they are repetitive.
We got some feedback from friends that having a visual clue for the exercises could make a big difference. Something like a video or animation would allow users to follow along without needing to read through the steps. I completely agree with this idea, as it makes the experience much more intuitive.
We are considering two approaches to add this feature. One idea is to create a static library of exercises with pre-recorded videos or animations. This would provide clear and consistent guidance but might feel repetitive over time. Another option is to dynamically generate animations or visuals from the exercise descriptions, as explained in this twit [1]
What do you think? Would having videos or animations improve the experience for you? And would you prefer a fixed set of ~100 exercises or dynamically generated visuals? Your input will help us decide how to move forward.
To figure out what topics ChatGPT users are likely searching for, I’d start by using proxy data sources like Google Trends or keyword research tools like SEMrush or Ahrefs. These tools can give you a decent idea of popular search topics around LLMs, AI assistants, or conversational tools—stuff that might overlap with what people ask ChatGPT. I’d also keep an eye on forums like Reddit or Quora, or even Twitter, where people tend to discuss these tools a lot. Spotting recurring questions or themes there could help you get a sense of what users are after.
Another option would be to look into Chrome extensions made for ChatGPT. Some of these plugins might collect anonymized data, and their descriptions or user reviews could hint at common patterns. That said, to get anything really useful, you’d probably need to either reach out to the authors of those extensions or create your own. If you go the second route, you could build an extension that improves the ChatGPT experience in some way while (with consent, obviously) gathering anonymized data on user queries and interactions. Just make sure it’s privacy-compliant and adds actual value to the user—otherwise, no one’s gonna use it!