Beyond the Wow Factor: Why We’re Waiting for z-Image’s i2i Model

The tech world is buzzing right now. The complete version of z-image has been released, sparking excitement about its raw generation capabilities. But while we’re all waiting for the next update to finally launch Image-to-Image (i2i), a lot of people are wondering if that delay just means the project is stalling.

But what if this wait isn’t about bugs or delays? What if it’s actually a sign that they’re getting their foundations right before releasing the final piece?

Distilled vs Foundation Models

To understand why we need to wait for i2i, you first have to look at the difference between «distilled» and «foundation» models. Right now, most of what’s out there is distilled—trained by larger models to be obedient, predictable, and safe.

Distilled models are great for productivity, but they’re boring for entertainment.

The complete foundation model that just launched is different. It has more internal capacity and less domestication. It’s designed to be expressive rather than rigidly filtered. When you look at the current hype cycle, it’s clear that people want this raw potential because they are tired of models that simply repeat styles or play it safe.

The Reality of Training LoRAs

While we wait for i2i, users are already experimenting with LoRAs. This is where the dynamics get interesting. There’s a distinct difference between using these tools for creation versus editing. For text-to-image generation, few-step LoRAs can be disastrous—they impose too much too early and block alternative paths.

However, when applied to image editing, they become incredibly useful. Because the composition is already fixed by the base image, a strong, short-impulse LoRA allows for rapid style transfer or mood changes without needing deep exploration.

This distinction explains why users are pushing so hard on the current T2I version. They’re using LoRAs as a crutch because they can’t yet leverage that powerful editing capability.

Why the Delay Matters

The delay in releasing i2i isn’t just about technical hurdles; it’s a strategic move to ensure quality. By holding back, the developers are avoiding the release of a tool that might encourage users to rely on aggressive few-step LoRAs as a substitute for actual editing.

When i2i finally launches, we won’t just be seeing another generation model. We’ll be seeing a shift in how these models interact with human intent—moving from pure imagination to practical, iterative refinement.

This entire process highlights that the current hype is just the beginning of a much more complex and rewarding development cycle.