Firefly's AI runs your whole workflow

PLUS: Why OpenAI killed Sora and DeepSeek V4 opens the frontier to everyone

Author

Gen AI Art
April 28, 2026 • Estimated Reading Time: 4 minutes

In partnership with

How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

Adobe just turned Firefly into a creative director — the new Firefly AI Assistant, now in public beta, orchestrates multi-step workflows across Photoshop, Lightroom, Premiere, Illustrator, and Firefly itself from a single conversational interface. Describe what you want to make, and the assistant coordinates execution across every app in the stack.

The launch also bundles in GPT Image 2, Runway Gen-4.5, Veo 3.1, Kling 3.0, and ElevenLabs Multilingual v2 — positioning Firefly less as a standalone image tool and more as the interface layer between creators and the entire AI model ecosystem. Whether this reshapes professional creative workflows or just adds a new layer of complexity is the real test.

Today in AI:

  • Adobe Firefly AI Assistant goes public beta — orchestrating all your Creative Cloud apps on command

  • OpenAI officially shuts down Sora after $1 million per day in losses

  • DeepSeek V4 open-sources a frontier model at a fraction of the competition’s price

What’s new? Adobe launched Firefly AI Assistant in public beta today — a creative agent that coordinates multi-step workflows across Photoshop, Lightroom, Premiere, Illustrator, and Firefly from a single conversational interface, with access to 60+ professional-grade tools across the Creative Cloud suite.

What matters?

  • The Creative Skills library pre-builds common workflows like batch photo editing, portrait retouching, mood board creation, and social asset generation — describe the outcome in plain language and the assistant handles every step.

  • Third-party model integration now brings GPT Image 2, Veo 3.1, Runway Gen-4.5, Kling 3.0, and ElevenLabs Multilingual v2 inside the Firefly interface, letting creators draw from multiple AI systems without leaving Creative Cloud.

  • The public beta is live now for Creative Cloud Pro and paid Firefly plan subscribers, with complimentary daily generative credits that refresh each day during the beta period.

Why it matters?

Creative tools have always operated in silos — switching between Photoshop, Premiere, Lightroom, and third-party AI models has meant constant app-switching and workflow fragmentation. Firefly AI Assistant puts one conversational layer in front of all of them, shifting the creative process from app-switching to intent-stating.

GUIDE

What’s new? OpenAI officially shut down Sora on April 26 — the app and web experience are gone, with the API following on September 24, 2026. The Wall Street Journal reported the service was losing $1 million per day, and professional adoption never reached sustainable levels in two years of operation.

What matters?

  • Despite massive initial hype, creative professionals didn’t integrate Sora into regular workflows — researchers identify a “counter-creative bias” in text-prompt video models, which are trained to reproduce familiar polished outputs rather than generate genuinely novel work.

  • OpenAI is reallocating compute to reasoning and agent capabilities, suggesting dedicated video generation infrastructure didn’t align with the company’s longer-term product direction.

  • The Sora API stays live through September 24, giving developers time to migrate to alternatives like Runway Gen-4.5, Kling 3.0, Veo 3.1, or Wan 2.7.

Why it matters?

Sora was the highest-profile bet that text-to-video would go mainstream for professional creative work — and two years in, it couldn’t hold users beyond the novelty period. With OpenAI’s resources redirected, Runway, Kling, Veo, and Wan now have clear territory to build the audience Sora assembled but couldn’t keep.

SPONSORED BY SPARROW

AI Entered the Sports World. The Investors Are Already In.

Every industry has its disruption moment. Intelligent performance training is having its right now.

Sparrow turns your smartphone into a real-time AI coach, analyzing every movement, every swing, every rep. No hardware. No expensive lessons. Just results.

85% revenue growth. 250K users. A $1 Trillion+ market.

Shares are $0.26. Early investors get +10% bonus. Round is almost closed.

The disruption is already happening.

Get in early: Invest in Sparrow now.

This is a paid advertisement for Sparrow's Regulation CF offering. Please read the offering circular at invest.sparrowup.com.

What’s new? DeepSeek released V4 — a 1.6 trillion parameter open-weights model with a 1 million token context window — pricing API access at $1.74/$3.48 per million input/output tokens, compared to $5/$30 for GPT-5.5 and $5/$25 for Opus 4.7.

What matters?

  • V4 ships in two variants: V4-Pro (49B active parameters) for frontier-level reasoning and V4-Flash (13B active) for high-throughput applications — both fully available to download and fine-tune on Hugging Face under an open-weights license.

  • The 1 million token default context is now standard across all DeepSeek services, enabling workflows that process entire creative codebases, long-form scripts, or large asset libraries in a single call.

  • The US State Department has accused DeepSeek of “industrial-scale” IP theft from American AI companies, ahead of a House bill that could add alleged distillation offenders to export blacklists.

Why it matters?

Open-weights frontier models at this price point lower the cost of building AI-powered creative tools significantly — from commercial image generation pipelines to video workflows that depend on underlying language models. The IP theft allegations add real uncertainty about long-term availability for Western developers, but for now the model is live, open, and cheap.

This massive rollout moves drone delivery from a novel experiment to a mainstream logistics operation for one of the world's largest retailers. The move accelerates the race for automated, on-demand commerce, pushing the entire last-mile delivery industry forward.

Everything else in AI

Wan released version 2.7 on Together AI, making the open-source text-to-video model available via free API — with outputs that rival commercial alternatives costing up to $200/month, built on Alibaba’s Tongyi Lab architecture.

Odyssey launched Odyssey-2 Max, its most powerful world model yet, generating physics-accurate interactive simulations continuously for over 120 seconds — the longest crash-free run achieved by any world model to date, with applications in gaming, film, and robotics.

Anthropic published results from Project Deal, a real employee marketplace where Claude agents negotiated 186 transactions worth $4,000+ — with Opus-backed agents earning sellers $2.68 more per item than Haiku-backed counterparts, and users unaware of the performance gap.

OpenAI launched Workspace Agents in ChatGPT, replacing custom GPTs with Codex-powered team agents that retain memory, integrate with Slack, and run on schedule — free until May 6, then credit-based for Business and Enterprise plans.

Essential AI Guides - Reading List:

Let us know!

Work with us

Reach 100k+ engaged Tech Professionals, Engineers, Managers and decision makers. Join brands like MorningBrew, HubSpot, Prezi, Nike, Ahref, Roku, 1440, Superhuman, and others in showcasing your product to our audience. Get in touch now →