Hello everyone. I’ve been a long-time reader of the discussions here, but this is my first time posting. I wanted to share my experience after my small marketing studio made the decision to switch our primary AI image generation tool from Midjourney to Adobe Firefly about two months ago. I tend to be quite methodical about these things, so I’ve compiled a detailed list of observations from our day-to-day use across various client projects. I’m hoping this might be useful for others considering a similar move, and I’d be very interested to hear if your experiences align.
Our main use cases involve creating supporting imagery for email campaigns, social media assets, and occasionally mockups for landing page concepts. We needed something that could integrate more cleanly with our existing Adobe Creative Cloud workflows and had stronger considerations for commercial safety.
Here are the most significant pros and cons we’ve identified so far:
**The Pros (What’s Working Well):**
* **Seamless Integration:** The direct integration into Photoshop (Beta) and Express is a massive workflow win. Being able to generate a background, select part of an image and extend it, or create a variation without constantly switching tabs and downloading/uploads has saved us a considerable amount of time.
* **Commercial Peace of Mind:** This was a major factor. The fact that Firefly is trained on Adobe Stock and public domain content means we can use the outputs for client work without the underlying worry about copyright infringement that always lingered with other tools. It removes a significant layer of legal risk assessment for our studio.
* **User Interface and Prompting:** The interface feels more intuitive to our designers who are already in the Adobe ecosystem. Features like the “Content Types” dropdown (Photo, Graphic, Art) and the “Style” presets provide a helpful structure for those less familiar with crafting elaborate text prompts. It has a lower initial learning curve for non-specialists on the team.
* **Transparency and Ethics:** The Content Credentials tag (even if it’s just a starting point) and the clear stance on training data align with our clients’ increasing requests for ethical sourcing. It’s a point we can mention in our process discussions.
**The Cons (Where We Feel the Friction):**
* **Creative “Spark” and Stylistic Range:** This is the most common feedback from our creative leads. While Firefly is excellent for generating realistic photos and clean graphics, it sometimes struggles with the more highly stylized, artistic, or conceptually abstract outputs that Midjourney seems to excel at. The “look” can feel safer, sometimes bordering on generic, which isn’t always what we need for standout campaign visuals.
* **Prompt Adherence and Detail:** We’ve found Firefly to be less consistent with complex prompts involving specific spatial relationships, multiple subjects, or intricate details. It often ignores or misinterprets finer points that Midjourney would handle more reliably. This leads to more generation cycles to get close to the desired result.
* **Speed and Generation Limits:** On busy days, the generation speed can drop noticeably, and we have to keep a close eye on the monthly generative credit allowance. With Midjourney’s subscription, the unlimited relaxed mode, while slower, felt less restrictive during high-volume periods. This requires more planning and allocation within our studio.
* **Lack of a True Community/Inspiration Hub:** A surprising downside has been missing the explorative aspect of Midjourney’s public gallery. Browsing what others created was a huge source of inspiration and a fantastic way to learn advanced prompting techniques. Firefly feels more like a solitary tool in comparison.
Our current stance is that the switch was the right operational and legal decision for our specific studio workflow, but it came with a creative trade-off. We now use Firefly for probably 80% of our work—especially where realism, safe commercial use, and Adobe app integration are key. For the other 20%, where we need a very specific, artistic, or unconventional style, we still maintain a single Midjourney account for those specialized tasks.
I’d be curious to know, especially from others in agency or studio settings:
* Have you found effective prompting strategies to push Firefly into more varied or artistic styles?
* How are you managing the credit system across a team? Any best practices for allocation?
* Does the integration with other Adobe apps like Illustrator or After Effects feel as mature as the Photoshop integration yet?
Thank you for reading such a lengthy first post. I’ve double-checked all these points with our team to ensure they reflect our collective experience.
~Heidi
I run a boutique agency focused on the hospitality niche, where we handle everything from customer lifecycle emails to promotional social visuals, and I've been testing both Midjourney and Firefly in production since each launched, using them weekly for client assets.
My comparison based on hands-on use across dozens of projects:
1. **Commercial Safety & Legal Peace of Mind:** Firefly's trained on Adobe Stock and public domain content, which means the generated images come with a commercial indemnity clause. For client work where you can't risk IP ambiguity, this is a concrete advantage. Midjourney's terms are murkier for commercial resale.
2. **True Workflow Integration vs. Standalone Tool:** Firefly inside Photoshop (Beta) is a genuine workflow shift, not just an API bridge. For example, generating a background directly onto a layer, or using the 'Generative Fill' brush to extend a client's product shot, cuts steps. Midjourney, via Discord or the web, remains a separate ideation step requiring manual import.
3. **Cost Structure for a Small Team:** Midjourney's basic useful plan is around $30/month for unlimited generations, which is predictable. Firefly costs via Creative Cloud All Apps ($60/month per user) plus generative credits; you get 100 monthly credits, and prompt-heavy days can burn through them. The effective cost per image is higher for us.
4. **Artistic Style Range & "Feel":** Midjourney, especially v5.2 and v6, still produces more stylistically varied, "artistic" results with a wider emotional range. Firefly's output is often more sterile, corporate-safe, and can struggle with specific artistic styles (like '90s anime poster'). For purely creative concepting, Midjourney wins.
I'd recommend Firefly for studios already embedded in the Adobe ecosystem doing client commercial work where legal safety and Photoshop integration are non-negotiable. If your priority is raw creative inspiration, speed, and lower fixed cost, Midjourney is still the stronger tool. Tell us how much your work involves existing Photoshop files and whether your clients ever question image provenance.
don't spam bro
Seamless integration with Photoshop is a powerful benefit, but the performance metrics are what really determine if it's a viable replacement. Have you done any timing comparisons on generation latency for similar prompts between the two systems? In our tests, the iterative workflow of generating, tweaking, and regenerating inside Photoshop suffered from a significant slowdown per iteration compared to Midjourney's dedicated channels, which compounds quickly on tight deadlines.
The other tangible cost is the credit system. While convenient, have you tracked your studio's actual credit consumption versus your old Midjourney subscription cost? For high-volume asset creation, the per-image cost can escalate, especially when you need multiple variations to nail a client's specific request. Midjourney's flat rate, for all its workflow clumsiness, at least provides predictable overhead.
Show me the benchmarks
You're absolutely right about the performance impact. The per-iteration latency in Photoshop is a real bottleneck for rapid prototyping. We found the same thing, and it forced a change in how we batch requests. We treat it like a slow database query now, sending multiple prompt variations in one go from a prepared list, then selecting and editing the best outcome, rather than live tweaking.
On cost, the credit system does require more discipline. For predictable, high-volume template work, it's less economical. But we've found it breaks even for our mixed workload because it eliminates the separate subscription and reduces time spent on manual compositing. The real cost variable is the designer's hourly rate saved versus the generation credits burned.
sub-100ms or bust
That batch request trick is smart, I wouldn't have thought of it. The comparison to a slow database query makes total sense.
So your trade-off is basically designer time vs. credits. How do you track that to know it's breaking even? Do you have a rough ratio you aim for, like one credit should save X minutes?
Ask me in a year
Great first post, and thanks for sharing your detailed list. That seamless integration you mentioned is a total game-changer for email marketing assets. I can't tell you how many times I've needed to tweak a background for a campaign hero image to get the right vibe, and being able to do it right inside Photoshop without a round trip is a massive time saver.
One caveat from our work with Klaviyo templates - watch out for texture and style consistency across a batch of images for a single campaign. Sometimes Firefly's output for similar prompts can vary slightly in lighting or texture, which you'll notice when you lay out multiple images in a single email. Midjourney felt a bit more predictable there.
How have you handled generating multiple images for a series, like a 3-part welcome email sequence? Did you find a trick to keep the style uniform?
Always A/B test.
Your second point on workflow integration is critical for evaluating the total cost of ownership. The time saved by avoiding the import-export shuffle with Midjourney directly impacts the designer's hourly cost you mentioned.
However, that seamless integration hinges on having an active Creative Cloud subscription. For a team already on CC, the incremental cost is just the Firefly credits. But for a studio considering Firefly without existing Adobe seats, the true cost includes the full CC subscription plus credits, which changes the ROI calculation significantly compared to a standalone Midjourney plan.
Buy once, cry once.
Batching prompts like slow queries is such a practical workaround. We started doing something similar, but found it requires a different kind of prompt engineering upfront - you have to anticipate the variations you'll need, which is almost like writing a small script.
It makes me wonder if Adobe will eventually add a local history or prompt queue feature to Photoshop's interface, since this seems to be a common adaptation. For now, I keep a text file open next to PS with my batch list.
Clean code, happy life
That prompt queue idea sounds like a feature they'll eventually charge extra for in a 'pro' tier. But honestly, it highlights a deeper issue with this whole batch-as-workaround approach. You're not just scripting prompts, you're forcing a deterministic, waterfall-like planning phase onto a creative process that's supposed to be iterative. The latency is so bad it's changing your design methodology.
We tried the text file batch list too, and it just became a graveyard of unused variations because you can't perfectly predict which direction the client or your own aesthetic sense will want to go next. So you either waste credits on pre-generated options that get discarded, or you accept the latency hit for the true iterative step, which defeats the whole point.
I'd be more interested in seeing someone hack together a local script that uses the API to pre-gen a cache of base images based on a project's mood board, then pulls them into Photoshop. Treat it like a local CDN for generated assets. That at least separates the slow generation cycle from the actual design session.
Your k8s cluster is 40% idle.
The seamless integration into Photoshop sounds like such a huge time saver. That's one of the main things that's making me consider a switch. Do you find that your whole team picks it up quickly, or is there a bit of a learning curve to the new workflow inside the Beta? Especially for people who weren't doing much with AI before?
Your point about the workflow becoming a deterministic waterfall is astute, and it mirrors a common failure pattern in distributed systems. You design for the cache hit, but the real cost is in the cache miss.
> local script that uses the API to pre-gen a cache of base images... Treat it like a local CDN
We built a version of this. The hidden cost isn't the script, it's the ongoing ops overhead. You need a scheduler, local storage management, and a way to tag and search the cache, which becomes its own brittle mini-application. The latency just moves from the designer's UI to the devops backlog.
The core issue is that the system's high tail latency forces you into predictive execution, which has a low success rate for creative tasks. Until Adobe improves their p99 generation times, any workaround, including a local cache, is just adding complexity to compensate for a fundamental platform SLA.
Latency is a liability
You mentioned commercial safety as a driver for the switch. That's a critical point everyone glosses over. Have you actually reviewed Adobe's indemnification terms? The integration is smooth, but your liability exposure changes completely. Midjourney's commercial terms are murky at best. Firefly's are clear, but that clarity comes with Adobe's own compliance overhead. Your studio now hinges on their internal governance staying tight. It's a different kind of risk.
Trust, but audit.
Exactly! The integration into Photoshop Beta is fantastic for those quick in-situ edits. For a lot of our email work, it lets us adapt an image to a specific layout block in minutes instead of jumping between tabs and apps. That said, I've found it really forces you to think in terms of layers and selections from the start - you almost have to plan your compositing before you even generate, which is a subtle but real mental shift for our designers.
Happy testing!
That seamless Photoshop integration you mention is the killer feature on paper. But I've been trying to untangle what it costs in practice.
The real pricing isn't just the credit pack. It's the mandatory Creative Cloud subscription as the entry fee. You have to track that fixed overhead against the variable credit burn. For a small studio, those seats are a major line item that doesn't scale down with a quiet month.
Did you bake the full CC subscription cost into your comparison? I've seen teams get excited about the integrated workflow, only to realize their effective cost per generated image spikes because they're now paying for Illustrator and a dozen other apps they don't use for this workflow. It turns the cost model from a pure usage-based one into a hybrid with a high fixed floor.
Exactly. The subscription is the anchor dragging down the unit economics.
The quiet month problem is real. You're not just paying for unused credits, you're paying for unused Photoshop. That's the catch with their 'seamless' integration - it only feels seamless because you've already bought the entire platform. And the monthly commitment means you can't just pause it between projects like you could with a Midjourney plan.
Has anyone actually tried calculating the effective cost per image with that fixed cost spread over their typical monthly output? Bet it makes the credit price look trivial.
Read the contract