The Render

Happy post-NAB hangover week, and this year’s show delivered: Blackmagic's DaVinci Resolve 21 public beta dropped with hundreds of new AI tools and a Photo page squarely meant to eat Lightroom's lunch, while Adobe countered with Premiere's editor-first Color Mode plus a Firefly AI Assistant that wants to drive your whole CC suite by prompt.

And for those of you still holding on to Sora remember that it goes dark on Sunday (04/26/26) — six months after the launch that had Hollywood talking, OpenAI is quietly winding the consumer app down and pivoting its compute elsewhere (Images 2.0 perhaps? See below).

And our lede this week: YouTube finally opened its deepfake detection tool to everyone in Hollywood, it’s not just for politicians and journalists anymore.

What Happened? YouTube is extending its likeness protection tool — previously limited to government officials, journalists, and political candidates — to actors, musicians, athletes, and creators industry-wide. Talent uploads their likeness, YouTube's system flags potentially infringing AI-generated content for takedown. Access doesn't require a YouTube channel. The platform says it's working with major talent agencies to refine the workflow.

Why Is This Important? This is the platform-level response to what Charles Rivkin at the MPA has been hammering for months — most notably after that Brad Pitt vs. Tom Cruise Seedance clip made the rounds in February. For commercial producers and studios negotiating talent deals, knowing YouTube now offers a working takedown mechanism changes the risk conversation around AI likeness use. It's not a solution to the broader deepfake problem, but it's a real tool on the biggest video platform in the world.

What Happened? Blackmagic's annual NAB reveal brought Resolve 21 into public beta — a free upgrade, as usual. Headliners include a brand-new Photo page with node-based stills grading (yes, RAW support, yes, Lightroom catalog import), plus AI tools like IntelliSearch for timeline content search, CineFocus for post-capture focal shifts, and facial tools that handle de-aging, beauty work, and feature reshaping.

Why Is This Important? This is Blackmagic's biggest swing yet at Adobe, with the Photo page openly targeting Lightroom while the AI tool sprawl encroaches on territory colorists and editors have been paying plugin vendors to cover. For colorists and editors already in the Resolve ecosystem, IntelliSearch alone changes how you navigate long-form work — natural-language search across objects, dialogue, and faces is the kind of quality-of-life upgrade that compounds across every project. And it's still a free update for Studio owners.

What Happened? Adobe's NAB reveal centers on Color Mode in Premiere — a grading environment built specifically for editors rather than colorists, now in public beta. Frame.io Drive lets teams work on cloud projects as if they were local. Firefly adds Kling 3.0 and Kling 3.0 Omni, and Adobe previewed Firefly AI Assistant — a conversational layer that orchestrates Photoshop, Premiere, Lightroom, and more from a single prompt.

Why Is This Important? Color Mode is the interesting one — Adobe openly admits it was designed for editors who do their own finishing, which describes a huge swath of commercial and content shops that don't have a dedicated colorist. If it genuinely simplifies grading without a Lumetri-level learning curve, Adobe just collapsed one of the most common friction points in small-team post production. Frame.io Drive, meanwhile, continues the drift toward cloud-native collaboration that's reshaping how distributed (read: remote) teams share and review, and it feels squarely aimed at LucidLink — which many thought might be acquired by Adobe. Apparently not.

What Happened? Anthropic rolled out Claude Design — a conversational tool for generating prototypes, decks, and visual assets — built on the new Claude Opus 4.7 model and powered by Canva's design engine on the back end. Available to Pro, Max, Team, and Enterprise subscribers. Outputs export to PDF, PPTX, standalone HTML, or straight into Canva for collaborative editing. Claude can read a team's codebase or design files to apply brand systems automatically.

Why Is This Important? For agencies, production companies, and even post houses pitching constantly, the pitch deck grind is real — and this is aimed squarely at it. More broadly, it signals where the AI assistant business is heading: away from text boxes and toward finished visual deliverables. The Canva partnership is shrewd — Anthropic gets visual output without building it, Canva locks in its position as the editing layer for AI-generated work…win-win.

What Happened? OpenAI released ChatGPT Images V2 on April 21 — the first of its image models with native reasoning. Headline gains: 2K resolution, up to eight coherent images from a single prompt with character/object consistency, real-time web search for reference, and meaningfully better text rendering and layout control (multilingual text, infographics, slides, maps, even manga). Advanced "thinking" features are paid-tier only; base model ships to all ChatGPT users.

Why Is This Important? The character-consistency-across-eight-images piece is the real story for anyone doing concept work, storyboards, or mood boards — previously you'd fight the same model all day for consistent characters across frames. Typography that actually renders readable legible text in multiple languages means AI-generated signage, end cards, and mock-ups stop looking obviously AI-generated. Plenty of commercial teams will quietly retire their "don't use AI for anything with text" rule this quarter.

NOTE: DALL-E 2 and 3 will be “retired” on May 12 (more on this soon).

What Happened? Midjourney pushed V8.1 Alpha less than a month after V8.0 landed to widespread complaints about flat, aesthetic-less output. V8.1 restores the V7-era look, stabilizes moodboards and style references, and makes HD (native 2K) the default — 3x faster and 3x cheaper than V8.0's HD. Standard resolution is 50% faster and 25% cheaper. Image prompts and image weights, broken in V8.0, are back. Upscalers and V8-era edit/inpaint/outpaint tools are next on the roadmap.

Why Is This Important? V8.0 was a rare stumble for Midjourney, and for anyone using it for concept development, mood, or look dev, V8.1 is the course correction that restores the tool as a reliable part of the pipeline. The cost drop matters — 2K as default means you're not paying a premium for usable-resolution output. If you'd abandoned V8 after the initial release, it's worth re-opening. Now if they’d only address spelling…

What Happened? xAI quietly dropped Grok 4.3 Beta on April 17 — no press release, just appeared in the model selector for SuperGrok Heavy subscribers ($300/month). The model adds native video understanding (analyze and reason about video content directly) and can generate formatted PDFs, PowerPoint decks, and spreadsheets inline. xAI has already applied the video capability to Grok Imagine, which now produces more coherent extensions from source clips by actually reading the original video context.

Why Is This Important? Native video understanding is the underrated capability here — a model that can watch a rough cut and give real feedback, or a shot and flag continuity issues, changes what AI can do in an editorial context. The $300/month gate keeps it in experimental territory for most of us, but the trickle-down to Grok Imagine's video extension tool shows up in actual output quality within days. If you trust xAI, this is worth watching once pricing normalizes.

What Happened? OpenAI is shutting down the Sora app on Sunday, April 26, exactly six months after a launch that had Hollywood genuinely rattled. The API follows on September 24. Peak users hit around a million; by early 2026 that had collapsed under 500,000 while the app reportedly burned roughly $1 million a day in compute.

Why Is This Important? Sora was the tool that triggered the industry's "are we all going to be replaced" panic. Its retreat is the clearest signal yet that consumer text-to-video, at least for now, is a compute sinkhole without a revenue model. The center of gravity shifts to Kling, Runway, Veo, and now Grok — tools built for working creators rather than casual slop.

NOTE: For anyone still holding Sora-generated assets, break out your thumb drives because the export window closes this Sunday.