What changed in Midjourney V8.1 and why it matters now
Midjourney V8.1 became the default model on June 10, 2026, replacing V8 as the model every prompt runs against unless you specify otherwise. Standard jobs render about 4 to 5 times faster than V8, native 2K HD images skip the upscale step entirely, and a refined Raw mode finally gives you direct control over the rendering pipeline. Most users have not adjusted their prompts to take advantage of this.
If your Midjourney output looks the same as it did three months ago, that is the reason. V8.1 reads prompts differently. It holds small details that V8 dropped, renders quoted text accurately, and lets you turn off Midjourney's signature stylising "auto-pilot" with a single flag. The cost stays the same. The output ceiling does not.
This is a hands-on guide to three V8.1 capabilities most users have not switched on yet: HD image generation, Raw mode, and quoted-text prompts. Each section gives you the exact parameter, the visual difference it makes, and the type of work it unlocks.
What are HD images in Midjourney V8.1?
HD images in V8.1 generate at roughly 2K resolution natively. Where V8 produced a standard 1024-pixel render that you then upscaled to 2K through a second job, V8.1 generates the high-resolution image in a single pass. The result is sharper textures, cleaner skin and fabric, and significantly better legibility for product shots and posters. Use the --hd parameter at the end of any prompt to enable it.
The practical difference is visible immediately on small details. Watch the eyes, jewellery, hair strands, fabric weave, and any printed text in the image. V8.1 HD holds these elements where V8 smeared them. For client work and any image that will be printed or projected larger than a phone screen, HD is now the baseline, not a premium upgrade.
Try this prompt right now:
--- A close-up portrait of a Hong Kong woman in her late 30s, soft natural window light, wearing a simple white shirt, photographed on a Hasselblad medium format camera, sharp focus on eyes, shallow depth of field, professional headshot quality --ar 4:5 --hd --raw
Run the same prompt with and without --hd on V8.1. The eye detail, hair texture, and shirt fabric will be visibly cleaner in the HD version. If you are still producing portrait or product work without it, you are leaving the upgrade on the table.
One caveat: HD jobs use more compute. A Pro plan that lasted you all month at V8 settings might run out faster with HD turned on for every job. Use HD for final outputs and Standard mode for exploration.
How does Raw mode change your output?
Raw mode strips Midjourney's default aesthetic processing. Every Midjourney version applies a signature look on top of your prompt, cinematic lighting, dramatic shadow, slightly stylised colour. Raw mode turns that off. What you describe is what you get. Use the --raw parameter to enable it. Raw is the difference between Midjourney interpreting your prompt and Midjourney rendering it.
The clearest test: run a simple prompt like "a photograph of a banana on a kitchen counter" with and without --raw. Without Raw, you will get a cinematic, magazine-style banana with directional lighting. With Raw, you will get something closer to a phone snapshot. That is the point. Raw gives you a neutral starting point.
When to use Raw mode:
--- Product photography that needs to match real catalogue shots, not look cinematic
--- Realistic documentary or editorial photography
--- Reference images you plan to edit further in Photoshop or another tool
--- Any project where Midjourney's default colour grade clashes with your brand
When NOT to use Raw mode:
--- Movie posters, dramatic concept art, or anything that benefits from cinematic styling
--- Quick mood boards where the default look is what you want
--- Stylised illustration that leans into Midjourney's aesthetic
The mistake most users make is leaving Raw off and then writing increasingly elaborate prompts to fight the default look. Turn Raw on first, then add only the styling you actually want. You will get there in fewer iterations.
How accurate is text in V8.1 prompts?
V8.1 finally renders quoted text in prompts with high accuracy. Street signs read correctly. Product labels are legible. Poster headlines display the exact words you specified. V8 was unreliable here, with typography often coming out as gibberish or scrambled letters. V8.1 closes most of that gap for short, clean strings.
The technique is simple. Put the text you want rendered inside double quotes directly in your prompt. Keep the string short, ideally under 25 characters. Specify the typeface direction. Place it on a clearly defined surface in the scene.
Try this prompt:
--- A vintage Hong Kong noodle shop signboard in red and gold paint, the signboard reads "金記麵家" in bold calligraphic Chinese characters, weathered wooden background, narrow alley setting, atmospheric photography, late afternoon light --ar 3:2 --hd --raw
V8.1 will render the Chinese characters with reasonable accuracy in most generations. V8 produced unrecognisable scribbles on the same prompt. For mixed English-Chinese signage, English alone, or short product names on packaging, the success rate is now high enough to use professionally with a quick proofread.
Limitations to know: Long sentences still degrade. Anything over 30 characters becomes a roulette. Body text in books, newspapers, or screens will still come out as filler-looking glyphs. Use quoted text for headlines and signage, not paragraphs.
How do you combine HD, Raw, and quoted text in one prompt?
The three parameters stack. You can use --hd --raw together with quoted text on the same prompt and the effects compound. HD gives you the resolution, Raw gives you neutral rendering, and quoted text gives you legible typography. This stacked combination is the V8.1 workflow most professionals are now defaulting to for production work.
The four-part V8.1 production prompt template:
1. Subject and context. Describe what you want and where it is. Be specific about objects, lighting, mood.
2. Camera and style anchor. Name the camera, lens, or photography style. This grounds the image in a recognisable visual language.
3. Quoted text (if needed). Add any required signage or labels in double quotes, short and clear.
4. Parameters. Append --ar (aspect ratio), --hd, --raw at the end.
Full template example:
--- [Subject], [environment], [lighting], [mood], photographed on [camera/lens style], [the sign/label reads "TEXT"], --ar 3:2 --hd --raw
Save this template. Adapt the bracket fields for each project. You will get to a usable image in one or two iterations instead of five.
What are the common mistakes that ruin V8.1 output?
Most V8.1 output problems trace back to four predictable mistakes. Each has a specific fix you can apply immediately. Adjusting these before generating saves entire batches of compute.
Mistake 1: Over-stuffing the prompt. V8.1 reads prompts more carefully than V8, so cramming 40 descriptive words confuses it. The model tries to honour everything and produces visual noise. Cut the prompt to under 25 words. Lead with the subject. Move secondary details to the back.
Mistake 2: Using V7-era style modifiers. Phrases like "hyperrealistic, 8K, ultra-detailed, masterpiece, trending on ArtStation" were V5-era prompt hacks. V8.1 ignores or actively penalises them. Replace them with one concrete photography reference instead, for example "shot on Leica Q3, f/1.7 aperture."
Mistake 3: Forgetting --raw on documentary work. If your image needs to look real, not cinematic, Raw is not optional. Without it, V8.1 will keep adding dramatic shadow and stylised colour. This is the single biggest reason professional photographers complain that AI images "look fake."
Mistake 4: Running HD on exploration jobs. HD doubles or triples the credit cost per image. For mood boards or initial concept exploration, run standard. Switch to HD only once you have locked the prompt direction. The "Run batch as HD" feature lets you re-render a seed-locked prompt in HD on demand.
How do you build a V8.1 workflow that scales?
A scalable V8.1 workflow has three layers: a template library, a parameter discipline, and a review filter. The template library is a saved set of your four-part production prompts for each project type. The parameter discipline is knowing when HD and Raw are on or off by default for that project. The review filter is the screening rule that decides which images go to final.
Step 1: Build five template prompts. Pick the five image types you produce most often, for example product hero shots, lifestyle scenes, headshots, hero banners, social tiles. Write a four-part template for each. Save them in a Notion page, a Google Doc, or a Claude project file.
Step 2: Lock parameter defaults per template. Product hero shots: --hd --raw. Mood boards: standard, no --raw. Social tiles: --hd, --raw off. Document this once per template so you stop deciding mid-flow.
Step 3: Set a review filter. For every batch of four images Midjourney generates, decide upfront whether you will keep zero, one, or upscale all four. This stops the trap of upscaling everything and ending up with three near-identical files cluttering your asset library.
Practitioners who treat Midjourney like a one-shot prompt machine plateau around basic competence. Those who build templates, lock parameters, and apply review filters move into the territory where AI image work feels like a controlled pipeline rather than a slot machine.
Conclusion: From scratching the surface to consistent output
V8.1's three upgrades, HD, Raw, and accurate text, only deliver if you actually flip the switches. Most users do not, and their output stays at V8 quality on a V8.1 plan. The template-and-parameter approach above turns Midjourney from a creative gamble into a tool that produces what you ask for, on the first or second try.
AI image generation feels intimidating because every release adds new parameters, new modes, and new edge cases. We know AI's cold edges. We know your real challenges. 28 years with UD, turning technology into a partnership with warmth. The point is not to memorise every flag. It is to build the smallest workflow that consistently produces work you can ship.
Ready to Build Your AI Image Workflow?
Knowing the parameters is one thing. Building them into a workflow that runs reliably across your whole team is another. UD's AI directory and consultation service map every tool to the specific business job it does best, and we'll walk you through every step, from tool selection to template setup and team rollout.
