Published March 4, 2024 · By CampaignsLive · Industry
Adobe’s Firefly platform had a difficult positioning problem through 2023 and the first months of 2024. The strategic case for Firefly was clean — the model was trained on Adobe Stock contributors who opted in to training compensation, which gave the platform an unusually defensible rights story relative to the broadly-trained alternatives. The aesthetic case was less clean. Firefly’s output, through its first two versions, sat noticeably below the quality bar that Midjourney v5 and v6 had established as the working ceiling for AI image quality.
The April 2024 Firefly Image 3 release closed most of that gap. The model’s output became competitive with the open-corpus alternatives on the aesthetic dimensions that had been the weak point. For brand teams that had been waiting for an ethically-sourced platform that did not require an aesthetic compromise, the release was the inflection point.
This is a working assessment of what changed, what it meant for the broader tooling market, and what brand teams should take from the shift.
What Firefly Image 3 actually delivered
Three specific improvements were visible in the April 2024 release.
The first was photorealism. Firefly Image 2 had produced output that was reliably useful for stock-replacement work but lacked the photographic plausibility that Midjourney was producing by late 2023. Image 3 closed most of that gap. The output crossed the threshold that the Pope-in-a-puffer-jacket moment from a year earlier had marked as the audience-relevant photorealism bar.
The second was prompt fidelity. The Image 3 model followed detailed prompts with substantially better fidelity than Image 2. Multi-element prompts that specified composition, talent direction, environmental detail, and atmospheric quality produced output that addressed each of the specified elements rather than collapsing into the model’s defaults. This is the dimension that DALL·E v3 had led on in late 2023; Firefly Image 3 reached parity.
The third was production-relevant resolution. Image 3 generated at higher native resolution than Image 2, with better detail preservation through Adobe’s enhanced upscaling. The output became usable for the production categories that Firefly’s previous versions had been below the bar for — full-page editorial, larger-format OOH, magazine spreads — without the artifact management that downstream upscaling typically required.
What the inflection meant for the tooling market
The release shifted the competitive dynamics in a specific way.
Through 2023, the brand-side AI tooling decision had a built-in trade-off. Teams could choose ethically-sourced training (Firefly) at the cost of aesthetic quality, or aesthetic quality (Midjourney, the better Stable Diffusion checkpoints) at the cost of training-corpus questions. The trade-off was real and unavoidable; brand teams that prioritized one had to accept the other.
The April 2024 release substantially reduced that trade-off. Firefly Image 3 was not the absolute aesthetic peak of the market — Midjourney v6 retained a small lead on certain registers, particularly the editorial and atmospheric work — but it was close enough that ethically-sourced training became a defensible choice for brand-creative production rather than a constraint that ruled the brand out of the highest-quality output.
The competitive response from the other platforms was visible within months. Mid-2024 saw OpenAI, Midjourney, and several open-source projects begin publicizing their own training-corpus arrangements with more specificity. The ethical-sourcing question, which had been Adobe’s sole differentiator, became a competitive dimension that the broader market began to compete on. The Firefly v3 release was, in retrospect, one of the moments when the trade-off between ethics and quality stopped being a trade-off and started being a base expectation.
What this meant for brand teams
Three implications became visible through the second half of 2024.
The “ethical” tier of AI tooling became a viable choice for production work. Brand teams that had been running Midjourney or DALL·E in production but were uncomfortable with the training-corpus questions had a credible alternative they did not have six months earlier. A meaningful share of brand-side production volume migrated to Firefly through the summer and autumn of 2024.
The aesthetic-only argument for the open-corpus tools weakened. A team that defaulted to Midjourney or DALL·E in 2024 could no longer point to aesthetic superiority as the sufficient reason. The decision now had to address the training-corpus question explicitly, which most teams had been avoiding.
Specialized brand-creative tooling kept its differentiation. Firefly Image 3, for all its improvements, was still a general-purpose image model. It was trained on Adobe Stock — broad, ethically-sourced, but broad. Tools trained narrowly on brand-creative corpora retained their differentiation in the categories where brand-specific visual register matters. The competitive landscape, by late 2024, had three tiers: broad-corpus open-trained tools (Midjourney, DALL·E), broad-corpus ethically-trained tools (Firefly), and narrowly-trained brand-creative-specific tools (CampaignsLive among others). Each tier served a different point in the brand creative production stack.
Where Firefly fit in the broader stack
A working pattern that emerged through 2024 for teams that adopted Firefly seriously:
- Firefly for the work inside Adobe. Generative Fill in Photoshop, Generative Recolor in Illustrator, the design-finishing operations where the deep Creative Cloud integration was the central value. The cleanest fit.
- Firefly for ethically-defensible standalone generation where Adobe Stock-derived training was an asset. Corporate communications, healthcare communications, regulated industries where the training-corpus question was a procurement requirement.
- Specialized brand-creative tools for production work in categories where the brand-specific visual register mattered more than the broad training corpus. Production-grade output for print, OOH, and high-volume brand campaign work.
The market did not, in the end, converge on a single tool. The Firefly v3 release did not create a dominant platform; it created a credible third option. The brands that have done well in the intervening period have been the ones that picked tools based on the specific point in their production stack each tool was best fit for, rather than trying to standardize on a single platform across the entire workflow.
What the inflection did not change
Two things remained roughly where they had been.
The first was that none of the major tools, including the post-v3 Firefly, addressed brand consistency at the depth that production work actually requires. Per-brand fine-tuning, reference image conditioning at production quality, and identity-stable variant generation remained, by mid-2024, the differentiator that separated brand-creative production tools from general-purpose image tools. Firefly improved on the general-purpose dimension; it did not move the brand-specific dimension.
The second was that print and OOH output remained largely a downstream workflow problem. Firefly v3 produced higher-resolution native output than its predecessors, but the CMYK conversion, ICC color management, and print-ready file preparation remained a Creative Cloud-side workflow rather than a Firefly-native one. For teams running significant print production, this was an acceptable arrangement (the work happens in InDesign and Photoshop anyway) but it was not a one-step pipeline.
The combination of these two — general-purpose ceiling raised, brand-specific and production-specific gaps unchanged — set up the next twelve months of brand-creative tooling evolution. The platforms that closed those two specific gaps became the production tools that brand teams converged on through late 2024 and 2025. For the broader trajectory, see From Generation to Production-Ready: The Quiet 2025 Shift.