Midjourney V7: Draft Mode, Voice Prompts, Video Gen & Why Users Are Divided

Midjourney is a 60-person bootstrapped research lab that has never taken outside funding, never run ads, and never offered a free tier. It has also produced the most aesthetically distinctive AI-generated images on the internet for three consecutive years — the kind that stop your scroll, make you look twice, and occasionally make professional artists furious.

V7 alpha launched on April 3, 2025 — exactly one week after OpenAI debuted its GPT-4o image generator that went viral for Ghibli-style photos. The timing was not accidental. Midjourney's response to a viral OpenAI moment was to drop its most ambitious model update in nearly a year, rebuilt from the ground up with a new architecture, a new prompting paradigm, and an entire new way to interact with the tool. The reception was complicated. Ethan Mollick, the Wharton professor and AI influencer, said he liked the new releases but found issues with V7 prompt adherence compared to V6 on identical prompts. Others loved it immediately.

This review covers every V7 feature in detail, what actually changed from V6, the honest user reception, video generation, full pricing, and what's coming next — including Niji 7 and whatever V8 looks like.

What Changed in V7: The Complete Architecture Rebuild

David Holz described V7 as having noticeably higher image quality with beautiful textures, and bodies, hands, and objects of all kinds showing significantly better coherence on all details. Those aren't PR claims — they're measurable. In practice, six-fingered hands went from "common" to "occasional." Complex objects like machinery or architecture maintain structural logic more consistently. A fashion photographer testing V7 observed: "Fabric rendering improved dramatically — I can see individual threads in knits where V6 gave me blurry texture."

The rebuild matters because it's not just about image quality. The new architecture is described as enabling fundamentally different prompting behavior — V7 may require different styles of prompting compared to previous versions, so experimentation is encouraged. Users who had perfected V6 prompt recipes found that those same prompts produced different results in V7 — sometimes better, sometimes worse — which drove the "feels like V6.2" complaints. The model is genuinely different; it just requires relearning what it's good at.

Draft Mode: The Feature That Changes Everything

Draft Mode generates images at ten times the speed and half the cost of standard rendering — Midjourney describes it as "so fast that we change the prompt bar to a conversational mode when you're using it on web." That last part is the real innovation. At standard Midjourney speed, iteration is slow enough that you plan your prompts carefully, hit generate, and evaluate the batch. With Draft Mode, the workflow becomes conversational.

Tell it to swap out a cat with an owl or make it nighttime and it will automatically manipulate the prompt and start a new job. Click draft mode then the microphone button to enable voice mode — where you can think out loud and let the images flow beneath you. For creative concept work, this is a fundamentally different experience. You're no longer submitting jobs — you're having a visual conversation.

Draft images are lower quality than full renders, but they're good enough to evaluate composition, lighting, and direction before spending credits on a full render. When you find a draft you like, you can select "enhance" or "vary" options to re-render it at full quality. The practical workflow: generate 20 drafts to find the direction, enhance the best one to full quality. Net result: more iterations, better final outputs, at lower total cost than V6 full renders.

Voice Prompting: Speak Your Ideas Directly

Users can now speak aloud to Midjourney's alpha website — provided they have a microphone — and the model will listen and conjure its own text prompts based on the spoken audio descriptions, generating images from this. Voice prompting is available inside Draft Mode via the microphone button in the prompt bar.

The practical value here is for people who think visually but struggle to translate their mental image into precise prompt language. Instead of spending 10 minutes engineering a prompt, you describe what you're imagining conversationally — "a moody photograph of a woman reading in an old library, warm amber light coming from one lamp, everything else in deep shadow, very cinematic" — and Midjourney interprets and builds the text prompt from your speech. For new users especially, this dramatically lowers the entry barrier. You don't need to learn Midjourney's prompt syntax to get good results from day one.

Personalization: The Feature Half the Community Loves and Half Ignores

V7 requires users to generate a new V7-specific personalization profile before using the model at all — a change from V6 where it was optional. Users rate 200 pairs of images, selecting which they prefer, and Midjourney builds a profile of their aesthetic taste. Every generation is then subtly tuned to match. The process takes 15–20 minutes.

User feedback is genuinely mixed. One designer says: "My personalized V7 consistently produces images closer to my brand aesthetic without extensive prompting." Another reports the effect was "too subtle to notice." The honest assessment: personalization works best if you have a consistent, strong aesthetic preference — professional photographers, brand designers, and digital artists with established styles see the biggest benefit. If you generate across many different styles and subjects, the personalization has less to work with and the effect is minimal.

The mandatory aspect frustrated experienced V6 users who just wanted to try V7 immediately. Personalization data from V6 is not automatically transferred — you must rate a new set of images specifically for V7. It's a legitimate friction point and the most common complaint in early reviews.

Video Generation (Launched June 2025)

Midjourney launched video generation in June 2025, turning static images into 5–21 second animated clips, described as great for social media and marketing. This was the feature the community had been waiting for since V6 — the ability to animate Midjourney's images without exporting to a separate tool like Runway or Kling.

The technical approach: Midjourney's video engine is built on the same training pipeline as its image models, prioritizing motion consistency and visual style preservation over raw resolution. Midjourney stated their video quality would be "10X better" than current AI video products — a claim that generated debate. Independent reviewers generally found the motion quality competitive with Runway Gen-3 on stylized/artistic content, weaker on photorealistic human movement. The sweet spot is exactly what you'd expect from Midjourney: visually stunning, artistically styled clips that extend the aesthetic of a static image into motion. For literal product demos or talking-head video, it's the wrong tool.

Midjourney V7 vs. V6.1: What Actually Improved

Area	V6.1	V7
Architecture	Iterative build on V5 base	Complete rebuild — "totally different architecture"
Hand / body coherence	Six fingers common	Six fingers occasional — meaningful improvement
Texture quality	Good	Noticeably higher — individual fabric threads visible
Prompt understanding	Strong	Smarter — less need for prompt engineering
Speed (Draft Mode)	~60 sec standard	~6 sec in Draft Mode — 10x faster
Cost (Draft Mode)	Standard rate	Half cost in Draft Mode
Voice prompting	❌ Not available	✅ Available in Draft Mode
Video generation	❌ Not available	✅ 5–21 second clips (June 2025)
Personalization	Optional	Mandatory to access V7
Text in images	~8% success rate	~10% success rate — still essentially broken
Discord requirement	Heavily Discord-dependent	Full web app — Discord entirely optional

Midjourney V7 Full Pricing (2026)

Plan	Monthly Price	Annual Price	Fast GPU Hours	Key Perks
Basic	$10	$96 ($8/mo)	~3.3 hrs / month	~200 images/month; commercial rights; no Relax mode; no Stealth
Standard	$30	$288 ($24/mo)	15 hrs / month	Unlimited Relax generations — sweet spot for power users; no Stealth
Pro	$60	$576 ($48/mo)	30 hrs / month	Stealth Mode (private gallery); 12 concurrent jobs; commercial rights
Mega	$120	$1,152 ($96/mo)	60 hrs / month	Stealth Mode; 12 concurrent jobs; highest Fast GPU priority; best for agencies

How to Use Midjourney V7 (Web App — Discord No Longer Required)

Getting Started:

1. Go to midjourney.com and sign up — Google or Discord account accepted
2. Subscribe to any paid plan (minimum $10/month Basic)
3. Complete V7 personalization: click your profile icon → "Create Personalization" → rate 200 image pairs (15–20 minutes) — mandatory to access V7
4. Head to midjourney.com/imagine — type your first prompt in the Imagine bar
5. Select V7 from the model version dropdown if it isn't already default

Using Draft Mode:

1. Click the "Draft Mode" button to the right of the prompt bar — it turns orange when active
2. A microphone icon appears — click it to enable voice prompting
3. Describe your image out loud or type — Draft Mode generates in ~6 seconds at half standard cost
4. Browse your drafts, click "Enhance" on any you like to re-render at full quality
5. Use conversational edits: "make it night," "swap the dog for a cat," "add fog" — Midjourney updates the prompt automatically and regenerates

Key Parameters for V7:

• --v 7 — explicitly sets V7 if it isn't your default
• --draft — activates Draft Mode explicitly in any prompt
• --style raw — disables V7's artistic interpretation; produces more photographic, literal output
• --ar 16:9 / --ar 1:1 / --ar 9:16 — aspect ratio for widescreen, square, or portrait
• --p — toggles your personalization profile on/off for a single generation
• --sref [URL] — style reference; upload an image URL to match its visual style
• --cref [URL] — character reference; upload a character image to maintain consistency across generations
• --stylize [0–1000] — controls how strongly Midjourney applies its own aesthetic interpretation (default 100; lower = more literal, higher = more artistic)

Midjourney V7 vs. Competitors (2026)

Factor	Midjourney V7	Adobe Firefly 3	GPT Image (OpenAI)	FLUX.2
Artistic quality	Best-in-class	Professional, clean	Very good, Ghibli-viral	Excellent, consistent
Text in images	~10% accuracy	✅ Strong	✅ Best-in-class	✅ Good
Commercial safety	⚠️ Active lawsuits	✅ Fully licensed training	⚠️ Lawsuits pending	⚠️ Mixed licensing
Video generation	✅ 5–21 sec clips	Limited	Limited	❌
Character consistency	⚠️ 3–5 generations before drift	✅ Strong	✅ Strong	✅ Best multi-reference system
Free tier	❌ Paid only ($10/mo min)	✅ Free with Adobe account	✅ Free in ChatGPT	✅ Open source
Speed iteration	✅ Draft Mode ~6 sec	Fast	Fast	Very fast

Niji 7: The Anime Model That Got a 2026 Upgrade

Niji 7 launched on January 9, 2026, bringing a major boost in coherency — addressing the most common complaint about Niji 6's tendency to produce anatomically inconsistent anime figures. Niji 7 is a dedicated model for anime, illustration, and Eastern art styles developed with Spellbrush, accessible on both the main Midjourney web app and a dedicated Niji Discord server. For creators working primarily in anime/manga aesthetics, Niji models have their own dedicated website and Discord server, providing a tailored experience for anime enthusiasts. Use --niji 7 in the Midjourney web app to access Niji 7 alongside V7.

The Honest Limitations

Text in images is essentially broken. Only a 10% success rate for readable text — V7 is not the tool to use when you need words in your image. Use DALL-E 3, GPT Image, or Ideogram and add text in post-production.

Character consistency beyond 3–5 generations drifts. A comic artist's assessment: "I can get 3–4 consistent shots of a character, then it starts drifting. Fine for concept art, frustrating for actual comic production." The --cref feature helps but doesn't fully solve this for extended sequences.

Active copyright lawsuits. Midjourney is involved in multiple lawsuits over alleged copyright violations, specifically for training models using web-scraped images without artist consent. Legal outcomes may define boundaries for training data legality. For brand work where legal clarity matters — Adobe Firefly's fully licensed training is the safer choice.

No free trial, no free tier. $10/month minimum to generate a single image. If you're comparing to ChatGPT's free image generation or FLUX's open-source weights, this is a real barrier.

V7 requires new prompting habits. Users with perfected V6 prompt libraries will find some recipes produce different results — sometimes better, sometimes not. V7 is an entirely new model with unique strengths and probably a few weaknesses, and may require different styles of prompting. Budget time to re-test your most-used prompts.

Frequently Asked Questions

What Is Midjourney V7?

Midjourney V7 is the latest flagship AI image generation model, released in alpha on April 3, 2025, and set as the default on June 17, 2025 — featuring smarter prompt understanding, stunning improvements in realism with richer textures and better hand/body coherence, Draft Mode (10x faster, half cost, with voice commands), and personalization on by default.

Is Midjourney V7 Free?

No — Midjourney has no free tier. Midjourney strictly avoids the free trial model to prevent bot abuse. The minimum plan is $10/month (Basic). Midjourney occasionally enables free trials for special events, but these are rare. Pricing remains unchanged: $10–$120/month.

What Is Draft Mode in Midjourney V7?

Draft Mode is half the cost and renders images at 10 times the speed. It enables a conversational mode on web — you can tell it to swap a cat for an owl or make it nighttime and it automatically manipulates the prompt and starts a new job. Click the microphone button in Draft Mode to enable voice prompting.

Is Midjourney V7 Better Than V6?

Objectively better on: hand/body coherence, texture quality, prompt understanding, speed (Draft Mode), and overall architecture for future updates. Contested on: whether the aesthetic output is better or just different. Users like Magnific AI founder Javi Lopez said V7 "feels more like V6.2" while others praised the improvements in aesthetics and workflow speed. New users almost universally prefer V7; experienced V6 users with established prompt workflows are more divided.

Does Midjourney V7 Have Video Generation?

Yes — Midjourney launched video generation in June 2025, turning static images into 5–21 second animated clips. Best for artistically styled clips and social media content; weaker than Sora or Runway for photorealistic human movement.

Do I Have to Rate Images to Use V7?

Yes — V7 requires users to generate a new V7-specific personalized style before using the model at all. This involves rating 200 image pairs, takes 15–20 minutes, and cannot be skipped. Your V6 personalization profile does not transfer automatically.

Does Midjourney Still Require Discord?

No — Midjourney introduced its web interface alongside V6.1 in August 2024, consolidating all tools — generation, editing, canvas, community browsing — into a single web app at midjourney.com. Discord is still supported but entirely optional. The web app is now the recommended access method.

What Is the Best Midjourney V7 Plan?

For most individual users: Standard at $30/month — the Relax mode gives you effectively unlimited generations with a short wait, making the ~600–700 usable images per month achievable without running out of Fast GPU hours. For commercial or agency work requiring private images: Pro at $60/month for Stealth Mode. Basic at $10/month is only practical for hobbyist-level use (roughly 200 generations per month maximum).