We're evaluating AI image generation tools for our marketing team. I've been testing InvokeAI because the self-hosted option seems good for data privacy.
But the UI feels awkward compared to something like Midjourney. Simple tasks like batch processing or adjusting parameters take too many clicks. Has anyone deployed InvokeAI in a larger team setting? Were there workarounds for the interface, or did you look at other self-hosted options?
Trying to figure it out.
Yeah, I ran into that exact issue. The UI felt slow for anything beyond a single image. Did your team end up scripting anything for batch tasks, or was the learning curve too steep?
Also curious, when you looked at other self-hosted options, did pricing for the alternatives get prohibitive for a whole marketing team?
The UI absolutely becomes the bottleneck with multiple users. We had our creative team try InvokeAI for two weeks, and the consensus was it's fantastic for privacy but clunky for daily work.
We ended up building a simple internal web app that uses InvokeAI's API directly. It just takes a text list and some preset style parameters, then queues the jobs. That bypasses the UI entirely for batch work. It's not trivial to set up, but if you have a platform engineering team, they can probably wire something up in a sprint.
Did you look into Automatic1111's Forge as an alternative? It's also self-hosted and the UI workflow is much more responsive for batch operations, in my experience. The trade-off is a steeper initial setup.
Automate all the things.
Bypassing the UI is the way. We did similar, wrapped the API in a Slack slash command. Marketing just types `/genimage a cute cat, style watercolor` and it dumps the result in a channel.
But that API approach exposes a new problem, the lack of queue management at scale. Invoke's API doesn't have priority levels or good job visibility. You'll need to build that too, or you'll just move the bottleneck from the UI to a black box where jobs vanish.