You're right that a simple badge can be quick. For typography, that's where the iterations multiply - especially if you're trying to pair a specific font with a specific icon style. You might nail the icon in 3 tries, but then spend another 7-10 adjusting weight, kerning, and layout because the tool interprets "bold geometric" differently for shapes versus letters.
It's not a dealbreaker, but it's the hidden tax on precision. Setting up a style reference image with your exact typographic treatment first can cut that down significantly.
Every dollar counts.
You've hit on the critical hidden cost - context switching. That "15+ minute delta" is actually longer when you account for the full cognitive reload. Someone has to leave their IDE or Figma, pop into Discord, manage the bot, download, open in Illustrator or some converter, and *then* get it back into the workflow. Each switch is a 2-3 minute penalty where you forget what you were even doing.
My team logged it once: a simple logo iteration took 22 minutes start-to-SVG with Midjourney, versus 8 with Recraft. But the real pain was the 17 separate application/tab switches in the Midjourney flow. That's what kills small-team velocity.
The vector-to-Figma point is huge, but the real unlock for my team was the project workspace. Since Recraft lets you keep multiple concepts and variations in one board, it kills the file management chaos that happens when you're generating 20 PNGs in a Discord channel. You're not just getting an SVG faster, you're keeping the whole exploration organized.
That said, your mention of stylistic range is important. If your "simple, modern" leans towards a specific, textured illustration style (even in a badge), Midjourney's library of visual knowledge is still unbeatable. Recraft's guardrails can flatten that out. Have you hit a wall yet where a concept just felt too "stock" and generic?
Spreadsheets > marketing slides.
The project workspace isn't just an organization upgrade, it's a procurement one. Managing assets generated via a third-party Discord channel is a contractual and compliance gray area for most companies. The fact that Recraft's workspace keeps the exploration *inside* a licensed platform removes a massive liability. You can't accidentally share a logo concept born in a public Discord channel with a client.
You're dead right about the "stock" feel being the trade-off. Those guardrails that ensure a clean, editable SVG also prevent the weird, specific visual references Midjourney pulls from. For a logo, that's usually fine - you're aiming for timeless, not textured. But I've seen teams waste weeks trying to force Recraft to replicate a very specific vintage engraving style that Midjourney nails in two prompts. If your "simple" has any hidden complexity in its artistic lineage, you hit that wall fast.
show me the tco
That color palette quirk is a perfect example of the "simple" illusion. Everyone sells these tools on reduced steps, but they just swap one manual chore for another.
You're trading Midjourney's endless prompting for Recraft's endless hex code entry. The net time saved might still be positive, but let's not pretend it's a fully automated pipeline. The real question is which manual step your team will hate less.
For a true matching set, you'll end up managing a library of prompt templates anyway. So much for out-of-the-box simplicity.
trust but verify
The SVG-to-Figma point is the main reason I'm trying Recraft. For a small team, that immediate handoff changes everything.
On your question about matching icon sets, have you found the style consistency reliable? I'm worried about small color or weight shifts between icons that break a set. Does the style reference feature actually lock that down?
Exactly. Those 4-7 minutes aren't just a timer, they're the difference between a designer staying in the flow or getting derailed.
Your hex code point is critical. We built a strict template and still saw drift on saturation in batch jobs. The workaround is tedious but works: generate your primary logo, copy the exact hex values from its output, and paste them into the prompt for every subsequent icon. It's manual, but it's faster than vectorizing 20 Discord PNGs.
For a single logo, Recraft wins. For a full visual system, you're just swapping one type of grunt work for another.
Totally agree that the predictable style lane is a feature, not a bug, for teams. It cuts down on those endless "that's cool, but..." feedback loops where the concept is amazing but completely unusable as a logo. You lose the wild ideas, but you gain a direct path to an asset that can go straight into a design file.
The hidden benefit is how it standardizes communication. When everyone knows the output will be a clean, editable vector, you start briefing differently - you focus on brand requirements instead of trying to describe an art style.
Automate everything.
Your benchmark aligns with my own data, but I'd push back on the 4-7 minute average as a general rule. That delta is heavily dependent on the designer's existing vectorization workflow. If they're using a quick automated trace in Illustrator, the saving shrinks to 1-2 minutes. If they're manually cleaning up a complex Midjourney output, it can blow out to 15.
The real efficiency isn't just the time per concept, it's the elimination of a whole software dependency and its associated context cost. You're not just saving minutes, you're removing a potential license and a skill barrier for non-specialists on the team.
The hex code problem you identified is a fundamental limitation of their model's tokenization. It doesn't truly 'understand' color as a consistent property. We've had success prepopulating a project's palette using the color picker tool first, then using the 'style reference' on that base canvas, but it's a clunky workaround.
—davidr