Piece No. 7: Visual Electric
Notes below.
I. Context
Visual Electric was founded in 2023. Image generation models, which have continued to improve in sophistication since, were launching rapidly - and a platform tailored to the designer use case had yet to be built. Visual Electric took a specific interest in supporting graphic & brand designers who are often responsible for creating editorial imagery, though it does offer integrations with Figma for more traditional product design use cases.
II. Choices
A | Center on a canvas
Given the focus on designers, the infinite canvas feels like a clear surface to build an experience around. The canvas encourages users to put generated images into context, allowing for fast comparison, and mirroring the expansive and non-linear creative process
This UI creates a subtle expectation to be able to change multiple elements on the canvas at once, which isn’t possible (at least today). Each image has to be individually adjusted
B | Encourage precise prompting
Visual Electric resists the common ‘chat’ interface other AI-powered tools leverage. There are fields to input prompts, but they are nested underneath different subfunctions or styling paths (e.g. “Retexture” which specifically adjusts the texture of a picture)
On one hand, this is a thoughtful approach that more overtly directs users towards steps they can take to adjust images and enables them to make these changes more accurately
On the other, it feels surprisingly disorienting given chat is the main form factor we expect to engage with today - though this anchor may evolve as LLMs improve (less inherent back & forth needed). It also takes away the feel of having ‘someone’ there to help - a byproduct of the chat / assistant UI - which can be unmooring
C | Expose models to users
This is a notable choice given the decision to abstract away other aspects of the “AI” experience from users. Perhaps it reflects the inputs designers want to have - freedom to customize creation, paired with more defined controls for editing
It also offers a gentle reminder of how many options exist when it comes to underlying models - and that real differentiation may exist where and how users choose to interact with the models (e.g. in an application like Visual Electric)
III. Conclusions
Visual Electric’s focus is clear and appreciated when the list of visual generative tools feels endless. This clarity also seems reflected in the aesthetic - light typography, ample visual space, a simple but striking color palette
Still, the tool feels unapologetically specialized - bells and whistles for an expert as opposed to a platform for the lay person to explore image generation - but again, that seems to be by design
Beyond positioning, meaningful product differentiation feels elusive. There are varied use cases across image generators - a Midjourney or Leonardo AI perhaps more purpose built for creation and iteration - but the gaps feel incremental. Maybe commoditization, even in the application layer, happens faster than anticipated