Hi friends. Coffee, and a topic that has been sitting in my drafts since the V8 alpha first showed up, because I wanted to actually live inside it before writing anything down. We have already covered which V8.1 styles are worth keeping, so today is not another style list. Today is the release itself: what V8 changed, what V8.1 fixed, and the one number nobody put in the headline that quietly changes how much your favorite workflow costs.
The short version is the one everyone latched onto. V8 is roughly five times faster than V7. A generation that used to take thirty to sixty seconds now lands in well under ten. That is genuinely lovely, and on its own it would be a great release. But speed was the headline, and the cost was the footnote, and the footnote is the part I want to talk about, because it is the part that hits your monthly bill.
What Actually Shipped in V8
V8 arrived as an alpha in mid-March 2026, built on a completely rewritten codebase rather than a tune-up of V7. That rewrite is why the speed jump is so dramatic, and it is also why a few familiar things behaved a little differently at first. The core wins are easy to summarize:
- About 5x faster standard generation. Standard jobs went from the half-minute-plus range to single digits. This is the change you feel every single time you press generate.
- Native 2K resolution. A new HD mode renders natively at 2K, so you are no longer chaining a separate upscale step just to get a usable resolution out of the model.
- Better text rendering and prompt understanding. The rebuilt model reads longer prompts more reliably and holds onto small details that V7 used to drop.
- Backward compatibility. Your existing V7 styles and moodboards still load, so nobody had to rebuild a reference library from scratch.
The Number That Was Not in the Headline
Here is the catch, and it is a real one. The premium features, the ones a lot of us treat as default, cost four times as much. Specifically, jobs using HD mode, the highest quality setting, style references, or moodboards currently run about four times slower than a standard job and burn four times the GPU minutes.
The headline is "five times faster." The thing that quietly reshapes your budget is "HD, style references, and moodboards cost four times as much."
Read that list again, because it is sneaky. Style references and moodboards are not niche power-user toggles. They are how most of us get a consistent look across a set. If your normal workflow is "lock a moodboard, generate in HD, refine," then your normal workflow just got four times more expensive per image in GPU-minute terms, even though each one finishes faster in wall-clock seconds. Fast and cheap are not the same axis, and V8 moved them in opposite directions.
| What you are doing | Relative speed | Relative cost |
|---|---|---|
| Standard V8 generation | About 5x faster than V7 | Baseline |
| HD mode (native 2K) | Slower than standard | About 4x the GPU minutes |
| Highest quality setting | Slower than standard | About 4x the GPU minutes |
| Style references and moodboards | Slower than standard | About 4x the GPU minutes |
The subscription price itself did not jump. What changed is how fast your included GPU minutes drain when you reach for the premium toggles. If you have ever hit your fast-hours wall halfway through the month, this is the release where you find out the hard way that your habits got pricier.
Then V8.1 Showed Up and Fixed the Mood
V8.0 also shipped with an aesthetic shift that the community was not in love with. People had spent a lot of time learning V7's look, and V8.0 rendered things differently enough to be jarring. Midjourney listened, and V8.1, which landed at the end of April 2026, brought the aesthetic back into the spirit of V7. If you tried V8.0 early, felt like your subjects looked subtly off, and quietly went back to V7, V8.1 is the version worth a second look.
V8.1 also chipped away at the cost problem, which is the part I care about most:
- HD is faster and cheaper. HD mode in V8.1 is about three times faster and three times cheaper than it was in V8.0. That does not erase the premium, but it meaningfully softens it.
- Standard got cheaper too. Standard resolution in V8.1 is roughly 50 percent faster and 25 percent cheaper than V8.0.
- More stable moodboards and style references. The features that cost the most are now more consistent, so you are less likely to burn four-times minutes on a generation that drifts off your reference.
- Image prompts and image weights are back. You can feed an image prompt and adjust its weight again, which matters a lot for anyone building a coherent set around a single reference.
- A new prompt shortener. Go over the prompt length limit and it kicks in automatically instead of just truncating your intent.
Native 2K is the headline feature here, generated at 2048 by 2048 pixels without a separate upscale pass. Combine that with the cheaper HD and the steadier references, and V8.1 is the version that finally makes the V8 generation worth the switch for everyday work rather than just experiments.
How I Would Actually Use V8.1 Without Torching My Minutes
- Explore in standard, finish in HD. Do your seed-hunting and composition rolls at standard resolution where it is cheapest and fastest, and only switch to HD once you have a generation worth committing to. Reaching for HD on every exploratory roll is how the bill runs away from you.
- Treat moodboards as a commit, not a default. Lock your moodboard once you know the look you want, rather than attaching it to every throwaway test. Each premium toggle is a four-times multiplier, and they stack with your own habits.
- Re-test your V7 prompts on V8.1, not V8.0. If your library felt broken on the early alpha, the V8.1 aesthetic return is the thing that brings your old prompts back to life. Do the comparison before you decide V8 is not for you.
- Use image prompts plus weights for set consistency. Now that they are back, an image prompt at a tuned weight is a cleaner way to hold a look across a series than stacking three style references and praying.
- Watch your fast-hours meter, not just the clock. Wall-clock speed went up, but GPU-minute cost on premium features went up more. The meter that matters for your wallet is fast hours remaining, not seconds per image.
The Honest Caveats
A couple of things are worth saying plainly. V8 is still labeled an alpha, and on the most complex, layered prompts it can lose the thread in ways the hybrid autoregressive competitors handle a bit better. The speed and resolution wins are real, but it is not magically smarter on every prompt. And there is still no public API, so if you were hoping to wire V8 into an automated pipeline, that door is not open yet.
The other honest note is the same one I keep coming back to with every flashy release: the headline number and the number that affects you are rarely the same number. "Five times faster" is the demo. "Four times the cost on the features you use most, softened by V8.1" is the line item. Both are true. Only one of them shows up on your statement.
The Bottom Line
For working AI artists, V8 is a real generational jump in speed and resolution, and V8.1 is the version that makes it livable, with a return to the V7 look, cheaper HD, steadier references, and image prompts back in the toolbox. Just go in with both numbers in your head. The speed is free. The premium look is not. Budget your fast hours like you mean it, explore cheap and finish expensive, and V8.1 will earn its place in your workflow without quietly eating your month.
I am going to go re-run a stack of my old V7 moodboards through V8.1 and see how many of them snap right back into place. If yours do, tell me which ones. Now, more coffee.