Midjourney V8.2 Is Default Now: What Changed and What's Still Missing

Midjourney shipped three version bumps in under five months this year. V8 landed as an opt-in alpha on March 17, 2026. V8.1 followed April 14 and took over as the default for every user on June 10. V8.2 then replaced it as default on July 24, 2026 (Midjourney version docs). That's a faster release cadence than the image-model space usually sees, and it's easy to lose track of what actually changed across three jumps versus what's still marketing language from the first one.

Versely runs Midjourney V7 for images and V7 Image to Video — no V8 endpoint yet. That's not a story about being behind; it's a reason to actually map what V8.2 improved against what V7 still does perfectly well, so you know whether you're missing anything and how to cover the one real gap: sound.

The V8 line, in order

Three releases, three different jobs:

• V8 alpha (March 17, 2026). The architectural reset. Midjourney rebuilt the generation pipeline from the ground up, moving off the older stack onto a GPU-native pathway. The headline gains: roughly 5x faster generation, native 2K resolution via a new HD flag, dramatically improved text rendering inside images, a more coherent multi-object scene composition engine, and a rebuilt style-reference system (WaveSpeed). This was opt-in, running on a separate alpha site while V7 stayed the default everywhere else.
• V8.1 (April 14, 2026; default from June 10). The stabilization pass. Midjourney's own framing was a return to a "more familiar" aesthetic in the spirit of V7, plus more stable moodboards and style references, a faster/cheaper HD mode, image prompts and weights, a prompt shortener, and a reworked Describe feature. This is the version that actually replaced V7 as default for the general population, ten weeks after the alpha shipped.
• V8.2 (July 24, 2026; current default). An aesthetics and personalization pass rather than a new architecture. Midjourney's documentation frames it as sharpening image quality and taste-matching: V8.2 outputs skew more creative, bold and sophisticated by default, and the Personalization system reads a user's aesthetic profile more accurately — especially for accounts with a large personalization point history — off a larger, improved pool of comparison images (Midjourney version docs).

The pattern across all three: the big structural jump was V8 alpha in March. V8.1 and V8.2 are refinements — stability, then taste — not repeats of the March leap. If you evaluated "V8" once back in March and moved on, you evaluated an alpha that's since been iterated on twice.

What actually changed since V7

Strip the version-number noise and three things are real, durable upgrades over V7's baseline: generation speed (the 5x figure is architecture-level, not a one-time promo), native 2K output without an upscale pass, and materially better in-image text rendering — historically one of Midjourney's weaker spots against dedicated typography-forward models. Prompt adherence on multi-object scenes also improved with the V8 composition engine, which matters more for complex briefs (three subjects, specific spatial relationships) than for single-subject shots, where V7 was already reliable.

What V8.2 specifically adds on top is subtler: it's a look and a personalization improvement, not a new capability. If your workflow doesn't lean on Midjourney's Personalization profiles, the V8.1-to-V8.2 jump is closer to a style refresh than a functional one.

V8 Video: still Pro/Mega only, still silent

Midjourney's first video model shipped as part of the V8 alpha program in April 2026, a few weeks after the image model's March 17 launch, and two things about it haven't moved since: it's gated to Pro and Mega subscription tiers, and it generates no native audio. Every V8 Video clip comes out silent — sound design is entirely a post-generation step, same constraint V7 Image to Video has always had on Versely.

That silence is the one gap worth planning around regardless of which Midjourney version you're on. The native audio glossary entry covers which current models generate sound alongside picture versus which ones — Midjourney's whole lineup among them — need it added afterward.

Where V7 still holds up

For anyone not chasing Personalization-profile refinement, Midjourney V7 — 10 credits a generation on Versely — remains a genuinely strong general-purpose text-to-image model: stylization quality, painterly and illustrative range, and the parameter controls (aspect ratio, style weight, reference images) that made Midjourney the default choice for a certain kind of creative-direction-heavy work are unchanged. The gap that matters most — 2K-native output and materially faster iteration — is real, but it's a speed-and-resolution gap, not a "V7 produces worse images" gap. For side-by-side testing against other current image models at the tier Versely actually runs, Midjourney V7 vs Ideogram 3 is the direct comparison.

Midjourney V7 Image to Video sits in the same position: solid cinematic motion and style consistency off a still image, silent by design, same as every V8 Video clip would be.

Worked example: scoring a silent Midjourney-style clip

Here's the actual workflow for closing the audio gap on a V7 Image to Video generation inside Versely, start to finish:

1. Generate your still in text-to-image on Midjourney V7 — get the composition and style locked before you spend on motion.
2. Send it to Midjourney V7 Image to Video to animate. You now have a silent clip with real camera motion and style consistency.
3. Prompt Versely's agent directly rather than hunting for buttons:

"Generate an ambient lo-fi track that matches the mood of this clip, then mix it under the video at low volume so the motion still reads as the focal point."

That routes to generate_music for the track, then attach_audio_to_video in mix mode to lay it under the footage — the same pairing covered in the add music to a video workflow. Because Midjourney's video output (V7 or, eventually, V8) is silent by construction, this pairing isn't a workaround for a temporary gap — it's the permanent second half of the pipeline for anyone building with Midjourney-family motion.

The actual takeaway

V8.2 is a real, cumulative improvement over V7 on speed, resolution and text rendering — three version bumps compounding since March, not a single flashy update. It is not a reason to abandon V7 output you already like, and it does not touch the one structural limitation that follows the whole Midjourney video line regardless of version: no native audio, ever, so far. Track the full model list if you want to know the moment a V8 endpoint lands here; until then, V7 plus a deliberate audio pass covers the same ground.

FAQ

Is Midjourney V8.2 a completely new model, or an update to V8.1?

An update, not a rebuild. The architectural jump happened once, at V8 alpha in March 2026. V8.1 was a stability and usability pass; V8.2 is an aesthetics and Personalization pass on the same underlying architecture. If you already evaluated V8.1's image quality, V8.2's delta is smaller than the version number implies.

Does any version of Midjourney generate video with sound?

No. V7 Image to Video and V8 Video are both silent — Midjourney generates picture and motion only, on every version shipped so far. Sound has to be added afterward, whether that's a generated track, licensed music, or a voice pass.

Can I use Midjourney V8 through Versely yet?

Not yet — Versely currently runs Midjourney V7 for images and V7 Image to Video. The model list reflects new endpoints as they're added, so that's the page to check rather than assuming a version number automatically means access.

Is V7 worth using now that V8.2 exists?

Yes, for most work. The V8 line's real advantages are speed, native 2K, and text rendering — a production and fidelity upgrade, not a taste upgrade. V7's stylization range and parameter controls are unchanged and still competitive against current-generation image models; see the V7 vs Ideogram 3 comparison for a direct read on where it stands today.