Has anyone else run into major consistency issues when switching Leonardo AI models mid-project? I'm working on a visual brand guide and was generating assets with Leonardo Diffusion XL, but needed to switch to Phoenix for a specific style. The output was completely different—colors, lighting, even the character proportions shifted.
It feels like a hidden vendor risk. We're evaluating Leonardo for ongoing production work, and this kind of unpredictability between models could really derail timelines.
**What I've tried so far:**
- Saving and reusing the exact same prompt and seed when switching models. The results are still wildly inconsistent.
- Adjusting the prompt strength and negative prompts to compensate, but it's guesswork.
I'm looking for a practical workaround. Are you:
* Locking in one model per project and never switching?
* Using a specific set of "bridge" parameters to smooth the transition?
* Exporting and processing images elsewhere to standardize the look?
Would love to hear how others are handling this, especially if you're using Leonardo in a professional pipeline where asset consistency is non-negotiable.
Ask me about my RFP template
Yeah, this is the exact reason we're still keeping a static style guide with hex codes and Pantone references on hand, even for AI-generated assets. The model switch is basically a hard reset.
> could really derail timelines
100%. We lock to one model per project after the style exploration phase. If we absolutely need a different model's look, we treat it as a new sub-project and generate a whole new batch with adjusted brand parameters from the start. It's more work, but less than fixing inconsistency later.
Have you found any external tool that's good for batch color correction? Trying to normalize outputs after the fact feels clunky.