Superstudio is a complex, modular creative tool with a lot of depth. That depth is what makes it powerful, but it was also killing first-session engagement. Session recordings showed a clear pattern: users would land on a blank canvas, hesitate, poke around, and leave. The onboarding tutorial had a 6% completion rate. Users who skipped it actually activated at a higher rate than those who completed it.
The data backed this up across the board. Canvas engagement dropped sharply after day one and continued declining. Only about half of users ever created a node, and just 43% started a media generation process. We were watching users churn before they ever experienced the thing that would make them stay.
This wasn't just a beginner problem. Even experienced users with local LoRA workflows and deep AI stacks struggled initially to understand how to use Superstudio. They stuck around because they believed in the output quality, not because the UX helped them succeed.
I started by asking a simple question: what do users want to make first? The data returned a simple answer.
I ran interviews with 20 users, 10 newer users and 10 long-term users. Despite totally different levels of product knowledge, there was a single recurring problem: presented with all possibilities, users do not know what to do first. A blank canvas is overwhelming, especially on a new surface. Tutorials, example canvases, and help docs were available, but I discovered a universal anxiety from users when starting on a blank canvas.
We knew from previous experiments that pre-loading the canvas with assets to experiment with introduced more confusion than it solved.
I also uncovered several secondary insights that shaped the final direction:
- Power users struggle too. Onboarding isn't just a beginner issue. Even advanced users found the initial experience confusing, and features we considered crucial to unlocking the product (like collections) remained unclear to many of them.
- Users want prompting help. Users are not confident in their ability to craft prompts or select models, leading to confusion about whether a poor output is their fault, Superstudio's, or the model's.
- Inspiration is valuable, but not as a first step. Users wanted content like galleries, featured work, and tutorials, but only after they'd already started creating. As a starting screen, it felt like a mood board, not a tool.
I ran usability tests on three iterations of this solution: a preloaded canvas, a simple blank canvas state with direct entrypoints to media creation, and a simple dashboard.
We landed on a simple design with four button options for Create Image, Create Video, Create Audio, and Explore All Tools. Instead of opening a model menu when clicking one of these buttons, the user is immediately taken to the canvas with one flow loaded using the most popular model for the media type chosen. This solution worked on a number of levels to solve the problem and a few unexpected ones:
- New users can start creating quickly without thinking about model choices
- Users are organically introduced to a core product mechanic that we struggled to communicate: flows are the core container needed to generate media. By clicking "Create Video" and immediately seeing a flow with an empty prompt field, new users rapidly understood
- We don't add any additional steps for veteran users. The screen acts as a shortcut to get where they know they want to go
- Kiko's prompt field persists at the bottom, giving users who prefer natural language a parallel entry point without cluttering the primary flow
A version of the design lives as an accessible home screen users can return to anytime for inspiration, templates, and model discovery.
This project, combined with related onboarding improvements across the product, drove a 20% increase in user retention and contributed to $100k in MRR growth. The solution reframed how we thought about the entire first-session experience: instead of teaching users the product, we got them creating immediately and let the product teach itself through use.