Midjourney V6.1 Default Model Update Breakdown: New Upscaler, Faster Speed, and More Accurate Text

So what exactly changed in Midjourney V6.1?

After Midjourney V6.1 launched, the most immediate changes are that it’s “more stable” and “faster.” It improves the coherence of human limbs and body structure by a noticeable margin, while reducing common pixel artifacts, making the image cleaner.

Also, Midjourney V6.1 is currently used as the default model, so you can experience the improvements without switching anything; if you want to confirm the model status, you can check /settings in Discord.

The new upscaler is more practical: textures, skin, and details hold up better

Midjourney V6.1 introduces a new upscaler (an update to the Upscale approach), focusing on improving “image and texture quality.” In practice, skin texture, material textures, and lighting/shadow depth are easier to get right, and the image no longer looks like it’s “smeared with a layer of oil.”

If you often make portraits, product images, or posters that need close-up detail, it’s recommended to generate a 2×2 grid with /imagine first, then Upscale the one that’s closest to your goal—it saves more time than repeatedly rerolling.

More accurate text generation: put the words you want to display in quotation marks

In the past, making “posters with text” in Midjourney often failed; V6.1 shows a clear improvement in text accuracy. A useful habit is to mark the text you want to appear with quotation marks in the prompt, such as “OPEN NOW” or “Coffee Festival.”

It still can’t guarantee perfect results every time, but in V6.1, quoted text is more controllable—good for getting the main headline working first, then fine-tuning layout and style.

Two tiers for speed vs. texture: Standard is faster, --q 2 is more detailed

Midjourney V6.1 speeds up standard generation tasks; the official figure is about 25%. If you’re racing to deliver options and need to iterate composition quickly, run drafts with the default quality first—it’ll go more smoothly.

When you need richer texture detail, you can try the --q 2 parameter. It takes longer (about 25% more time) to “carve out texture,” but it may sometimes slightly sacrifice visual coherence—better suited for a small number of key images before finalizing.

Quickly try the new features: one prompt is enough to verify the improvements

In Discord, go to your usual channel and enter /imagine. Start by testing Midjourney V6.1 with a “subject + material + lighting + lens” structure, for example: close-up portrait, skin details, shallow depth of field, studio soft light.

Then do two comparisons: one with normal generation, and one with --q 2; then use Upscale to compare texture differences. This way you can quickly judge, for your subject matter, whether V6.1 is “faster and more stable” or “more detailed but more costly.”