Midjourney releases V8.2 Preview with --preview and 24x faster --sref random drafts

• Midjourney opened V8.2 Preview behind a --preview prompt flag, according to midjourney's launch post, while midjourney's reply confirming it is live says people could start using it immediately.
• The other concrete ship was speed: midjourney's launch post says big batch draft mode now works with --sref random, which lets users explore style directions 24x faster than before.
• Midjourney is already steering testers toward a specific combo. In midjourney's personalization reply, the company tells people to try V8.2 with personalization and mood boards.
• Early workflow chatter from ai_artworkgen's tip turned the update into a practical loop: generate a 24-image draft grid, pick favorites, then use Vary > Subtle for high resolution versions.
• The preview is also surfacing adjacent experiments. chrisfirst's promotable timeline demo shows a timeline-like interface for morphing prompts across generated frames, which points past still-image prompting into sequence design.

You can see the official 24-image tiger grid in midjourney's launch post, compare it with chrisfirst's V8.2 portrait test, and jump to lloydcreates' sports fashion experiment for a more stylized use case. There is also a neat workflow wrinkle in ai_artworkgen's draft-mode tip, plus chrisfirst's promotable timeline demo hinting at a more direct way to steer transitions.

Midjourney shipped V8.2 as an opt-in preview, not a full model switch. The whole entry point is simple: append --preview to the end of a prompt, as midjourney's instruction reply puts it.

That matters because the company is still framing the model as in-progress. In midjourney's direction reply, it says the team is still working on V8.2 and hopes users like where it is going.

The most useful mechanical change is not a new aesthetic claim. It is the faster search loop around --sref random in big batch draft mode.

The workflow that emerged around the update is straightforward:

• Turn on Draft Mode, per ai_artworkgen's tip.
• Generate the 24-image grid with randomized style references, as shown in midjourney's launch post.
• Pick the cells worth keeping.
• Run Vary > Subtle to turn those selections into high resolution outputs, according to ai_artworkgen's tip.

That makes V8.2 feel less like a single-image model upgrade and more like a faster style-finding harness.

Midjourney is explicitly pairing the preview with its existing steering tools. In midjourney's personalization reply, it tells testers to combine V8.2 with personalization and mood boards.

That is a quiet clue about what the company wants people to evaluate first: not raw prompt adherence in isolation, but how the new preview behaves when fed a stronger visual preference signal.

The first public tests are already split between portrait work and more synthetic fashion-sports compositions.

The contrast is useful. chrisfirst's post leans into cinematic portrait lighting, while lloydcreates' follow-up test thread pushes V8.2 toward glossy product forms, reflective landscapes, and editorial-feeling character shots.

One extra experiment slipped into the conversation around the preview. In chrisfirst's promotable timeline demo, a video shows a prompt bar being dragged across a sequence so the image changes over time.

chrisfirst's promotable timeline demo

The post asks whether people would want a "promotable timeline like this," which is a small but concrete signal that Midjourney-adjacent creators are already thinking beyond single prompts and toward editable progression across frames. That is new information the V8.2 preview itself does not spell out.