Case01
Two AI ads win at Cannes, both built on Kling — the maker breaks down why
PJ Ace, who has spent two years building ads with Kling, flagged that two AI-made ads just won at Cannes — both built on Kling — and posted a breakdown of why it is a bigger deal than the headline. Awards move AI film from novelty to accepted craft: a Cannes win is the industry conceding that an AI-built ad can meet the bar top agencies compete on, which reframes the client conversation from can-you to why-not-you. For a studio the takeaway is to treat award-winning AI work as sales collateral and reverse-engineer it — what held (concept, casting, edit rhythm) versus what the model did — so you can quote the same ceiling to brand clients instead of leading with cost.
Source PJ Ace @PJaccetturo · 2026.07.12 ↗
Industry02
Adobe makes Firefly conversational: one chat runs the whole creative workflow
Adobe put a Firefly AI Assistant into public beta that runs an entire creative workflow through conversation — start from a brand idea, pick a name, palette and aesthetic, generate logo directions, then keep refining by chat instead of jumping between apps. This is the incumbent adopting the same agentic, describe-and-refine interface the AI-native tools have been racing on, and Adobe brings the one thing the startups lack: a captive base already living in its apps. For teams the read is that the conversational-creative interface is becoming table stakes, not a differentiator — so a studio competes on taste and finished quality downstream of the chat box, and should watch whether Firefly pulls casual brand work in-house before it reaches an agency.
Source Umesh @umesh_ai · 2026.07.12 ↗
Technique03
Precise camera-and-action control, prompted into Seedance 2.0
TechHalla shared a Seedance 2.0 result with what he calls total camera and action control, published with the full prompt. The signal is that control — separating camera move from subject action and specifying both precisely — is becoming a prompt-craft skill rather than a model limitation, which is exactly the axis where a shot succeeds or reads as generic AI drift. For teams this belongs in the same capture-recipe library as period looks: bank the phrasings that reliably split camera motion from action for this model, because that decoupling is what our previs layer enforces structurally and what a text-only workflow has to earn through wording.
Source TechHalla @techhalla · 2026.07.12 ↗
Tool04
Luma leans on Skills — a reusable creator capability as the unit of work
Luma showcased a shot made with a named Luma Skill by creator Eli Coleman — a reusable capability packaged and credited to its author. Skills as a distributable unit are showing up across platforms now, which reframes value from one-off outputs to codified, shareable method — the same bet Secret Sky makes by releasing selected methods as Skills. For a studio the implication is that a signature technique is an asset, not a trade secret to hoard: packaging a repeatable look or move as a named Skill compounds reuse across projects and doubles as marketing, provided the taste and judgment behind it stay yours.
Source Luma AI @LumaLabsAI · 2026.07.12 ↗
Case05
A full history documentary, cut by an agent (invideo agent one)
TechHalla made a full documentary of Columbus’s first voyage using nothing but an agent — invideo agent one — and documented the whole process, framing it as a change in how history gets recreated. It extends the director-agent thread from single clips and sitcoms to long-form nonfiction, where continuity, pacing and factual framing across many shots are the hard part — the first sign the format reaches documentary length, not just hero reels. For production teams the useful move is to probe where an agent-cut long-form actually breaks — narrative coherence, source fidelity, transitions — because those failure points are the editorial and QC layer a self-hosted pipeline should own rather than hand to the agent.
Source TechHalla @techhalla · 2026.07.13 ↗
