Midjourney makes V8.2 its default image model

• V8.2 is now Midjourney's default image model, and midjourney's launch post frames it around aesthetics, personalization, and image quality.
• Personalization is the sharpest creator-facing change: DavidSHolz's post singled it out, and DavidSHolz's image-pool reply said the new image pool is much larger.
• The preview build moved fast: midjourney's preview reply said the latest preview had only been tested for a day or two before release.
• Creators were already using V8.2 as a production ingredient, from a Midjourney image into a Hailuo, PixVerse, and CapCut fantasy sequence in DrSadek_'s workflow to a Seedance 2.0 griffin video with a Midjourney 8.2 environment in MayorKingAI's workflow.

The official V8.2 post says low-quality random outputs should be dramatically reduced and personalization should understand taste better, especially for profiles with lots of ratings. The Version docs quietly confirm the switch: V8.2 became the default on July 24, 2026, after V8.1 held the default slot from June 10 to July 23. An office-hours recap says the earlier --preview build was an approximation running on an existing V7 personalization profile, which explains why preview-era prompt shares may not perfectly match the default model.

Midjourney made V8.2 the default model and described the release as focused on "aesthetics, personalization, and image quality" in its official announcement. The model's new house style is described as more "creative, bold, sophisticated, edgy and fresh."

The Version docs now list V8.2 as the current default and say users can still set versions manually with --version or --v on web or Discord.

The release is really about personalization. DavidSHolz's post told users the new model is "especially" good there, while DavidSHolz's image-pool reply said the image pool for creating profiles is much larger now.

Midjourney's Personalization docs describe the system as a style assistant trained from images users like or select. V8.2 adds four practical changes around that flow:

• Better taste matching, especially for profiles with many ratings, according to the official V8.2 post.
• A larger and improved image pool for creating V8.2 personalization profiles, per the same post.
• Profile use through --p, profile IDs, and personalization codes, as described in the Personalization docs.
• Stylize as a control for how much personalization is applied, with values from 0 to 1000 and a default of 100, according to the docs.

Creators had been prompting V8.2 with --preview before the default switch. carolletta's question captured the simple usage pattern, and midjourney's preview reply said the latest preview had only been tested for a day or two.

A few mechanics are now clear from official docs and replies:

• V8.2 is the default, while older versions remain selectable with --v #, according to the Version docs.
• HD images are available at 2048px, matching midjourney's 2048 reply and the docs' V8.1 and V8.2 compatibility chart.
• HD costs 1.3 minutes of GPU time versus 0.8 minutes for SD in the Version docs.
• HD aspect ratio is capped at 4:1, while the general V8.1 and V8.2 max aspect ratio remains 14:1 in the same chart.
• --exp is listed in the V8.1 and V8.2 compatibility chart, and midjourney's exp reply said "exp is new too."

DrSadek_ used one Midjourney 8.2 image as the anchor for a short fantasy sequence, then moved through Hailuo Minimax Hub, PixVerse CPP, and CapCut. The starting image in DrSadek_'s anchor-image post was an old man on a misty coastal cliff holding a glowing cage or lantern.

The workflow detail that matters: DrSadek_'s reply says the backstory was not prewritten. The video storyline was written while creating the clip in Hailuo Minimax Hub.

MayorKingAI used Midjourney 8.2 differently: MayorKingAI's griffin workflow says the environment was designed in Midjourney 8.2 preview, with GPT Image 2 and Seedance 2.0 in Leonardo, music from Suno, and editing in CapCut.

The strongest prompt shares were not generic model tests. They were style systems.

• GTA loading-screen art: MayorKingAI's GTA style used --sref 7941867634 --preview.
• Cosmic vertical posters: carolletta's Electric Dreams prompt used --chaos 45 --ar 9:16 --raw --stylize 600 --weird 25 --preview.
• Infinite sci-fi space: carolletta's Infinite Library prompt reused the same parameter stack with a library concept.
• Neo-noir blue studies: ai_artworkgen's Feeling Blue prompt shared a token list built around saturated cobalt, blue-hour cinematography, cross-processed color, and Lynchian atmosphere.
• Aquatint prints: AllaAisling's aquatint prompt used a fill-in template with subject, two colors, rosin ground, hand-wiped plate tone, and deckled paper edges.

V8.2 was not the only Midjourney item in circulation this week. DavidSHolz's extra-detail link pointed users to more detail, and the same office-hours recap says V9 is already in training.

That recap also says Midjourney's editing models are training and should move into testing with Guides and Moderators soon. Its three listed priorities were finishing V8.2, continuing the next major version, and rethinking the image creation process.