The highly anticipated Midjourney V6 has officially landed, bringing forth a slew of enhancements in visual coherence, language comprehension, photorealism, and text generation. While not all of Midjourney’s staple features are available in this initial alpha release, the core image generation capabilities have seen substantial improvements.
The visual upgrades in V6 are truly remarkable, with enhancements spanning aesthetic quality, textures, lighting, and intricate details. The introduction of new prompting guidelines emphasizes the importance of cleaner and more specific prompts. V6’s sensitivity to language nuances marks a significant leap forward, opening up fresh creative possibilities for users. In order words – if you’re a seasoned prompt engineer used to working in V5, get ready to relearn everything you thought you knew about prompt engineering.
Highlights of Midjourney V6
- Realistic Imagery: V6 outshines its predecessors in generating more lifelike images, boasting increased accuracy in prompt following and handling longer prompts.
- Improved Coherence and Knowledge: The model exhibits enhanced coherence, coupled with a broader understanding of various prompts.
- Text Generation Debut: Midjourney V6 introduces text generation capabilities, a highly anticipated feature. While occasional spelling variations may occur, it represents a substantial improvement from the previous version.
How do I use it?
You can access V6 one of two ways: Turn it on as your default from the /settings menu or include the parameter --v 6.0 at the end of your prompt.
Exploring Midjourney V6 Features
Text Generation
Text generation is a significant addition to V6 we’ve all been waiting for. While it is not quite perfect, it can understand and generate short lines of text fairly well. Remember to place your intended text in quotation marks. Have a look at some examples below:
Prompt: A photograph of the words “Hello Winter!” written in white on a coffee shop window, warm lighting, winter vibes, cozy atmosphere. --ar 3:2 --v 6.0 --style raw
Prompt: A photograph of an alternative girl with vibrant pink hair holding a chalk board with the words “Pink Horn” written on it. --ar 3:2 --v 6.0 --style raw
Improved Upscalers
Midjourney V6 has introduced two new upscalers: “Subtle” and “Creative” modes with a 2x resolution increase. You’ll see these new options under your image once you click the standard U1, U2, U3 or U4 button.
Original Image:
Take a look at the upscale comparison below. Subtle on the left, Creative on the Right.
I honestly don’t know which one I prefer. I’m perfectly happy with the original upscaled image, but I admit that both of the new upscalers give the final image an extra little oomph. I think between Subtle or Creative on this particular piece, I prefer the Subtle Upscaler because it looks more natural. But I do like the sharper look of the eyes in the Creative upscaler. Play around with your results (keeping in mind that these upscalers are slower and cost more of your GPU minutes).
New Midjourney V6 Prompting Style
V6 introduces a different approach to prompting, emphasizing explicitness and avoiding unnecessary “junk” terms.
From the Midjourney Development team:
- V6 is MUCH more sensitive to your prompt. Avoid ‘junk’ like “award winning, photorealistic, 4k, 8k”
- Be explicit about what you want. It may be less vibey but if you are explicit it’s now MUCH better at understanding you.
- If you want something more photographic / less opinionated / more literal you should probably default to using
--style raw - Lower values of
--stylize(default 100) may have better prompt understanding while higher values (up to 1000) may have better aesthetics
The following features are supported at launch: --ar, --chaos, --weird, --tile ,--stylize, --style raw , Vary (subtle) ,Vary (strong), Remix, /blend ,/describe (just the v5 version)
These features are not yet supported, but should come over the coming month: Pan, Zoom, Vary (region), /tune, /describe (a new v6 version)
Prompt: Wildlife photography of a black panther drinking water from a creek. --ar 3:2 --v 6.0 --style raw
Prompt: Drone landscape photography of a glacier in Iceland --ar 3:2 --v 6.0 --style raw
Although I don’t think this feature has changed drastically, below is a --stylize comparison of the same prompt.
Prompt: A sassy black cat driving a moped on a high speed chase. –ar 3:2 –v 6.0
V6 vs V5.2
Comparing Midjourney V5.2 with V6 reveals distinct stylistic differences. V6 prioritizes realism and detailed imagery, making it suitable for photorealistic images. In contrast, V5.2 excels in creating creative graphics and illustrations with a stronger emphasis on aesthetics. In each of the examples below, I used the same prompts in V6 as in V5.2, specifying only the aspect ratio.
Keep in Mind
- This is an alpha test; regular and unpredictable changes are expected.
- Version 6 is slower and more expensive in GPU hours than Version 5 but optimizations are in progress.
- Anticipated enhancements include improvements in speed, image quality, coherence, prompt adherence, and text accuracy.
I can’t wait to see what the next updates will bring. Until then, I’ll be over here relearning how to prompt all over again.
Stay tuned for more updates and prompt articles!
Pingback: How to Generate Multiple Subjects in Midjourney V6 (Alpha) - Pink Horn