AI by Day, Visual Storytelling by Night

By day, I work on AI solutions for clients in the ecommerce space, but by night I work on AI solutions for visual storytelling. By visual storytelling I mean the focus is to tell an entertaining and engaging story, with meaning. Some may call it filmmaking, but that feels ostentatious. I am more interested in creating a YouTube series of episodes over a long period of time, allowing me to make releases, than a single feature film.

I started out with 2D animation in Adobe Character Animator, moved on to 3D animation in Unity and then NVIDIA Omniverse, but have more recently jumped on the Generative AI bandwagon using ComfyUI with a combination of free open source models and paid services. But the original domain name stuck - ordinaryanimator.com.

Building to learn

One of the reasons I am building it is to learn. Coding tools like Cursor and Claude Code etc. make cutting code much faster. That has allowed me to explore experiences combining AI agents with user experiences. I don't believe that all application interfaces should be a chat. So how to design systems, with good user experiences, to combine both? That has been the fun part of the journey for me. Trying things out, discarding approaches that felt clunky.

A chat that spans the whole app

So in ordinaryanimator.com (not open to the public!) I have the main traditional web app interface, a side panel chat with text and voice conversations, with deep integrations between the two. The UI side invokes chat sessions instead of pointing at help documentation, and includes "review my data" for insights. The chat side can update the underlying data structures directly, and navigate the user around in the UI. I tried chat experiences embedded in the UI, but chats often span multiple pages. It just felt wrong limiting chat to a single page. For example, "Go through all characters and scenes they appear in and change any mention of HARRY into SAM." The UI updates showing the user all the changes made on the screen while it performs the higher-order (multi-step) operation.

Online courses in the age of AI

But along the way I discovered something else. My YouTube feed has lots of videos talking about "online courses in the age of AI are not doing as well - information is now cheap!" You can just fire up ChatGPT or Gemini etc and get solid advice back. (Okay, usually solid advice - I always push back a bit to make sure it can justify what it says. On top of that, most people do not finish courses.) So a few "gurus" are recommending instead approaches such as cohorts. Not 1-on-1 coaching, not online courses, but group coaching with a cohort of students that talk and work with each other. The cohort serves to encourage people to keep going and complete their work. It feels more like a team event instead of a solo mountain climb. It still includes online content, but it is a more holistic approach.

But it made me think about my project and the integration of AI. The way my application is built is it asks for lots of different types of information. The goal is to get the user to think about their story from multiple aspects. Think about the mannerisms, quirks, preferences, strengths and weaknesses, dislikes, areas of growth, and so forth. It can feel like "yet another field?", but the reason is to force the user to consciously think it through, adding more depth to the story they later compose. (I don't use AI to write the story. I use AI to explain, educate, and review what a human writes.) AI can then cross check storylines and plots and timing sequences to make sure a series does not introduce inconsistencies. My target audience is hobbyists. For myself, I have a full time job. I want to tell stories on a YouTube channel in my spare time. I forget the details, so I want detailed note taking to act as a long term reference to make sure I don't introduce mistakes.

From documentation to teacher

I started having AI describe the purpose of a field, what the user should think about. That is effectively the same as documentation. But as I introduced AI review of user input to a field, it became more. They effectively get scored by a teacher as they progress. Documentation becomes targeted at the exact use case the user is looking at, because it has all the information available. It knows the characters, their motivation, the storyline.

I think this approach is potentially really useful. It is more than just teaching - it includes an immediate, targeted feedback loop. The user does not just consume material - they act upon it immediately, they get feedback immediately. So while I built the software as a product to create videos telling engaging stories, it may be it is more valuable as a platform to teach people how to write stories. Users don't passively watch online content then try to repeat, not knowing if they are doing it right. They get taught concepts with immediate feedback they can learn from, instantly.

I may never open ordinaryanimator.com up as a product, but it has been an interesting journey regardless. I think there are still a lot of unknowns about the combination of UX best practices and AI powered interfaces. I can see interesting years ahead for a while yet!