Midjourney V8 Alpha launched on March 17th, 2026. It generates images roughly five times faster than V7, renders text inside images that you can actually read, and outputs native 2K resolution without upscaling. It's available on alpha.midjourney.com, not in Discord, and some of the new flags cost 4x more GPU time. Here's everything we've learned using it for client work over the past three weeks.
What actually changed from V7
Three things matter for ecommerce work. Everything else is noise.
- Text rendering works. Put text in "quotes" inside your prompt and V8 renders it legibly. Product labels, packaging copy, signage, book covers. V7 could not do this reliably. V8 can. This alone changes what's possible for mockups.
- Native 2K with
--hd. V8's HD mode renders every pixel from scratch at 2048x2048. No upscaling artifacts. The output is clean enough for print. The trade-off: it costs 4x GPU time and doesn't support Relax mode yet. - Prompt fidelity. V8 follows complex, multi-element prompts much more faithfully. Specific colour palettes, spatial arrangements, lighting conditions, material textures: these used to get partially ignored. Now they render with noticeably higher accuracy.
The V8 parameter reference
These are the flags we use daily. Copy this block and keep it next to your prompt doc.
# V8 Alpha parameters (as of April 2026)
--v 8 # Use V8 model (required on alpha.midjourney.com)
--ar 1:1 # Aspect ratio. Common: 1:1, 4:5, 16:9, 9:16
--hd # Native 2K (2048x2048). 4x GPU cost.
--q 4 # Extra coherence pass. 4x GPU cost. Stack with --hd.
--stylize 200 # 0-1000. Default 100. Sweet spot: 100-400.
--chaos 20 # 0-100. Variation between outputs. Low = consistent.
--weird 0 # 0-3000. Experimental distortion. Keep at 0 for product work.
--p # Apply your Personalization profile (trained on your likes).
--raw # Less "Midjourney style", more literal interpretation.
--no shadows # Negative prompting. Exclude elements.
# Text rendering (NEW in V8):
# Put any text you want rendered in "double quotes" in the prompt.
# Example: a jar with a label reading "Wild Honey"
# Cost warning:
# --hd, --q 4, and --p each cost 4x GPU time.
# Stacking --hd --q 4 = 16x. Use sparingly.
Step-by-step: your first product shot in V8
Write the prompt like a photography brief
The biggest shift in V8 prompting is away from keyword lists and toward natural language. Write it like you'd brief a photographer: subject, setting, lighting, mood, camera.
PromptProduct photography of a hand-poured soy candle in a matte
black ceramic jar, placed on a raw linen cloth. Soft natural
window light from the left, shallow depth of field. Warm,
minimal, editorial. Shot on a 50mm lens. --ar 1:1 --v 8
Key insight: don't use the words "photorealistic" or "photorealism". They make V8 produce a digitised, uncanny-valley look. Instead, use "product photography", "studio lighting", or name a specific lens. The model understands intent.
Add text to the product
This is the V8 feature that changes ecommerce mockups. Put the text in quotes.
PromptProduct photography of a matte white ceramic candle jar
with a minimal black label reading "WILD HONEY" in clean
sans-serif type. Placed on a marble surface, soft studio
lighting, no shadows. --ar 1:1 --v 8 --stylize 150
V8 won't get it perfect every time. Expect 2-3 generations before the text is fully clean. Short text (1-3 words) works best. Full sentences are still unreliable.
Shoot at 2K for print-ready output
If the image is going on a physical product, a pitch deck, or a large hero section, use --hd.
Lifestyle flat lay of a luxury reed diffuser set, three
bottles in amber glass with "SCALE HOUSE" embossed on the
label, arranged on a dark walnut table with dried eucalyptus.
Overhead shot, natural daylight, editorial styling.
--ar 4:5 --v 8 --hd --stylize 200
Cost reality: --hd costs 4x GPU time. On the Basic plan ($10/month, ~200 images), that's roughly 50 HD images. On the Standard plan ($30/month, ~900 images), that's 225. Budget accordingly.
Build lifestyle scenes for ads
This is where V8 earns its keep for us. A single lifestyle shot for a Meta ad used to require a photographer, a stylist, and a half-day shoot. Now it's a prompt and sixty seconds.
PromptA woman in her 30s sitting cross-legged on a linen sofa,
holding a ceramic coffee mug, morning light streaming through
sheer curtains. Warm, calm, aspirational. The room is minimal
with muted earth tones. Shot on a Canon EOS R5 with an 85mm
f/1.4 lens, shallow depth of field. --ar 9:16 --v 8
For ad creative, we generate 6-10 variations and pick the best two. The --chaos flag helps here: set it to 30-50 for more variety between outputs.
Use negative prompting to clean up
The --no flag tells V8 what to exclude. Essential for product shots where you need a clean background.
Product photography of a white cotton t-shirt folded neatly
on a light grey concrete surface. Overhead shot, natural
daylight from the left, minimal shadows. Clean, editorial,
high-resolution. --ar 1:1 --v 8 --no people hands mannequin
wrinkles tags logos watermark text
What we exclude by default for product shots: --no people hands mannequin watermark text shadows. Add or remove based on the brief.
The prompt template
This is the template we copy-paste and modify for every client brief. It follows the structure that produces the most consistently usable output.
# Scale House Midjourney V8 product prompt template
#
# [SHOT TYPE] of [PRODUCT DESCRIPTION], [SURFACE/SETTING].
# [LIGHTING DESCRIPTION]. [MOOD/STYLE WORDS].
# Shot on a [CAMERA + LENS]. [DEPTH OF FIELD].
# --ar [RATIO] --v 8 [FLAGS]
#
# Example:
Product photography of a 250ml amber glass bottle with a
black dropper cap, placed on a raw concrete slab with a
sprig of rosemary beside it. Soft studio lighting from
above and slightly left, gentle shadows. Clean, minimal,
apothecary aesthetic. Shot on a 90mm macro lens, shallow
depth of field. --ar 1:1 --v 8 --stylize 200
--no text watermark people hands
What V8 still gets wrong
Honest notes from three weeks of daily use:
- Hands are better, not fixed. V8 produces fewer six-fingered horrors than V7, but if your prompt involves someone holding a product, budget for 3-4 generations to get clean hands.
- Long text is unreliable. One to three words on a label renders well. A full sentence on a poster still garbles. Workaround: generate the image without text, then add it in Photoshop or Figma.
- HD mode is expensive.
--hdat 4x GPU cost burns through your monthly allocation fast. We use it for hero images and pitch decks only. Social and ad creative stays at standard resolution. - No Relax mode yet. Every V8 generation uses Fast time. If you're on the Basic plan, you'll hit your limit in a day of heavy use.
- Personalization profiles from V7 mostly work. The
--pflag carries over but some profiles produce subtly different results. Re-train if something looks off. - It's alpha. The model is only on alpha.midjourney.com. Not in Discord, not in the main app. The interface is functional but things break occasionally. Save your prompts externally.
Prompts to try right now
Five prompts you can paste into V8 today. Each one is designed for a common ecommerce use case.
Hero banner (desktop)Wide cinematic shot of a woman walking through a sunlit
lavender field carrying a woven basket, golden hour, warm
colour grading, editorial fashion photography. Shot on a
35mm lens. --ar 21:9 --v 8 --stylize 300
Product grid (square)
Minimalist product photography of a small glass jar of
raw honey with a wooden dipper, white background, soft
even studio lighting, no shadows, clean and commercial.
--ar 1:1 --v 8 --no text labels logos people
Instagram Story (vertical)
Close-up of hands pouring coffee from a ceramic pour-over
into a handmade mug, steam rising, morning kitchen light,
warm tones, cozy and intimate. Shot on iPhone 15 Pro.
--ar 9:16 --v 8 --stylize 100
Meta ad creative (4:5)
Lifestyle product shot of a luxury scented candle burning
on a bedside table next to an open book and reading glasses,
soft warm lamp light, evening atmosphere, calm and premium.
Shot on an 85mm f/1.4. --ar 4:5 --v 8
Packaging mockup with text
Product photography of a kraft paper coffee bag with a
clean white label reading "MORNING RITUAL" in minimal
black sans-serif type. Placed on a dark slate surface
with scattered coffee beans. Soft directional light.
--ar 1:1 --v 8 --hd --stylize 200
Sources
Back to blog