Advertising

AI Pulse Labs and collaborators shared a computer-vision MVP that estimates billboard attention from pedestrian counts and head orientation. The prototype is still an internal experiment, but it could help teams test attention-focused measurement instead of raw impressions.

Runway introduced Agent 2.0 to turn a prompt into marketing briefs, creative assets, and performance analysis across platforms and markets. Try it if you want campaign planning and iteration inside one workflow instead of separate tools.

Runway released ad localization that turns a single image into market-specific variants across languages. Use it to keep multilingual ad production inside one creative tool instead of rebuilding assets for each region.

A builder demoed a UGC system that automates A/G steps, drafts posts, and keeps a side editor for manual control. It matters for repeatable ad production, but prompt tuning, account warm-up, and publishing results were still being worked through.

Creators on X reported recent payout cycles falling to €0 or about $95 even after millions of impressions, with some tying the drop to Nikita Bier's spam-account cleanup. Track payout eligibility and revenue rules closely if you depend on X funding generation costs; creators still say the system is opaque.

A creator walkthrough showed OiiOii Viral Clone ingesting a source ad, generating shot analysis, previewing its understanding, and then swapping characters, products, and styling while keeping the original pacing. That matters for UGC teams because it turns one proven ad structure into multiple branded variants without a manual rebuild.

Pika rolled out an MCP bundle for creator growth with Persona Builder, Content Director, Viral Hook, and Fix My Look. The release packages audience and presentation tasks as reusable agent skills, so creators can try the workflow across more posts and profiles.

Multiple creator demos showed Adobe for creativity in Claude handling template swaps, recolors, animation, and asset actions inside a single chat, with a second walkthrough covering skills-based setup. It matters because branded social production can stay inside Claude instead of bouncing between apps, though today’s evidence comes from sponsored creator demos rather than an Adobe launch post.

CapCut opened its first global AI festival with Film, Series, Creative, and Commercial tracks plus $200,000 in prizes. The event matters because AI-assisted work now has a cash-backed festival path tied to CapCut Video Studio, so creators can submit outside isolated prompt contests.

Creator posts showed FalcoCut generating influencer-style videos from templates with no prompt writing, followed by a short walkthrough of the selection flow. The workflow lowers the entry point for batch UGC production, but the evidence comes from one creator demo set.

Runway said token consumption rose 50% in six weeks, power users 140%, and enterprise NDR hit 300%. Those figures, plus $3K ad-remake claims and a nearly 100-startup Builders cohort, put concrete adoption numbers behind AI video.

Higgsfield showed Claude-driven brand identity, app screens, and motion-video generation inside its MCP, along with marketing skills and a Virality Predictor. The demos point to faster ad production; try the presets if you want to generate UGC-style variations at scale.

A new residency short, Steven Bakes a Cake, credits InVideo Agent One for story beats, felted animation, voiceover, sound design, and edit, and a separate creator used it for an Anti-Hero short. The posts add concrete narrative-film examples beyond last week's ad showcase; one of the pieces is a paid partnership.

Creators published a fully agent-made clothing-brand ad and a separate animated short built with InVideo Agent One for scripting, storyboards, generation, and editing. The examples matter because they extend InVideo's earlier one-prompt ad claim into public creator work, though the strongest demos come from partnership content.

Creators shared systems that scrape competitor ads, build inspiration vaults, generate static and video variants, and publish them straight into Meta campaigns. The workflows matter because they turn marketing into a repeatable testing loop, while adjacent examples show the same stack can be pushed into deceptive fake-character product videos.

Pika released a four-skill bundle for Claude via MCP covering brand creation, app screens, product sizzle videos, and founder videos. Demo posts show a single App Store link can generate polished promo assets inside one chat workflow, so founders can test it for faster launch content.

Runway opened ticket sales for its AI Film Festival in New York, led by a pre-show conversation with Ron Howard and a slate of 10 finalist films. It matters because the event packages AI film alongside advertising, design, new media, gaming, and fashion rather than a single-tool showcase.

Creators posted end-to-end ads made in InVideo Agent One, including a 30-second Alpine A390 spot and a full Luna Yoga retreat promo from a single prompt. The demos deepen earlier claims that Agent One can handle image generation, video, voiceover, sound, and assembly inside one flow.

Adobe Firefly AI Assistant entered public beta with creator demos focused on automatic crop, framing, and sizing changes from one image across multiple platforms. That turns a previously announced resizing feature into a live assistant workflow inside Firefly.

Promptsref showed a canvas that chains concepts, visuals, and video shots from storyboards into cinematic brand ads. A follow-up demo used the same canvas to adjust camera angle, so watch it as a production workflow rather than a one-off prompt share.

Fabian Stelzer showed a Glif agent that recreates reverse-poem ads and generates the copy, clips, voiceover, music, and final cut inside one workflow. Follow-up posts pointed users to a reusable prompt and another same-day Glif spot, making the format repeatable.

Luma Agents posted new examples for turning briefs and benefit hooks into UGC-style ad assets, while creators shared active boards for related ad builds. The evidence points to repeatable low-polish brand ad generation instead of single polished hero spots.

Multiple creators posted low or inconsistent X payouts, including one $62 period on 3.7 million video views, while Logan K said a flagged case was sent to the team and an appeal path should exist. The reports suggest impressions alone may no longer explain earnings, and suspected reposter flags can suppress revenue without clear diagnostics.

Magnific opened a €10 million program for European marketing teams that bundles 30% off Business credits, workflow playbooks, and team training through June 30. Consider it if you want lower access costs plus onboarding support for agencies and in-house teams.

Creators showed Agent One generating a 30-second commercial from one instruction and one product image, while another thread used it to build a film bible, storyboards, assets, and animation setup. The posts describe automation across voice, music, scenes, and edits, with final assembly still cleaned up in Premiere.

Creators are using Seedance 2.0 to turn reference images and storyboard sheets into ad spots, indie clips and realistic UGC from a single product shot. Use a first-frame pass followed by an animation pass to keep consistency and test variants faster.

Creators posted Seedance 2.0 workflows that turn one product shot, storyboards, or children's art into UGC ads, travel vlogs, and storybook clips. These runs matter because they document repeatable prompts and reference setups that others can try for production work.

Several X creators posted metrics showing 85-95% reach declines after viral posts and tied the drop to Phoenix-era trust scoring and paid-post handling. The evidence is still partly anecdotal, so watch disclosure and penalty transparency before drawing firm conclusions.

Elon Musk says X will publish the Phoenix feed system and future changes on GitHub with monthly release notes. Creators are already analyzing reply, dwell, native media, and negative-signal weighting in the Grok-assisted stack.

Higgsfield launched Supercomputer, a cloud agent that routes work across GPT 5.5, Opus 4.7, Gemini, Seedance 2.0, Nano Banana, and built-in campaign tools. It matters because brand memory, connectors, and generation now sit in one creative workspace instead of separate model tabs.

A walkthrough showed OmniSocials importing past posts, drafting platform-specific copy, and scheduling publishing across 10 networks for $10 per workspace through Claude. The setup bundles calendars, formatting, analytics, and scheduling into one chat-driven social stack.

New examples and docs showed Claude Code's /goal feature being used for lint-and-test passes, SEO audits, content queues, and outreach research with explicit proof and limits. That matters because the feature is moving beyond coding checks into repeatable creator and marketing workflows, but vague goals still drift without measurable end conditions.

Luma Agents added Kling Omni as a generation option and paired the integration with demos that carry a reference moodboard through to finished ad visuals. The update gives creators another video model inside Luma's existing campaign workflow.

Higgsfield opened an experimental preview that scores uploaded clips for viral potential and hook strength, then demoed it inside an MCP loop with Ad Reference. The preview currently does not consume credits, giving creators a measurable feedback layer for short-form ad iteration.

Higgsfield says Ad Reference MCP lets agents ingest winning video ads and generate new variants around the same patterns. The launch lands alongside Luma campaign builders and creator reports of Claude-and-Seedance phone-demo pipelines, pointing to repeatable ad iteration systems rather than one-off prompts.

TRIBE v2 is being used to score up to 100 short-form ad variants before paid testing, using attention, motion, emotion, memory, and scene cues. It predicts brain response instead of CTR, and the CC-BY-NC license limits agency use.

Hermes Agent playbooks now show the agent writing its own skill file and running a 7-agent SEO loop claimed to get posts indexed in under 14 days. That makes Hermes look more like a reusable operating layer, with Claude Code as the execution handoff.

Creators are using Seedance 2.0 prompts to fake handheld UGC ads, paparazzi-style crowd scenes, and shaky-phone footage with blocked sightlines and flash spill. Similar realism demos in ImagineArt and Kling suggest this look is becoming a repeatable workflow.

Youraipulse and Amir Mushich posted side-by-side TRIBE v2 analyses against YouTube and TikTok data, including one YouTube example with 130% average viewed and 81.5% retention. The posts push the tool from cut scoring toward prediction-versus-performance claims, but the evidence is still creator- and vendor-reported.

A free local tool built on Meta TRIBE v2 now scores uploaded videos with predicted response graphs, edit suggestions and multi-cut comparison. That matters because creators can test alternate edits before publishing, though the release is still framed as a community prototype rather than an official Meta editing product.

Higgsfield launched Canvas, a node-based workspace for repeatable content pipelines from brainstorming through final cut. Posts around the launch also pointed to new MCP hooks, tying the canvas approach to ad automation and team production workflows; test the graph if you need a structured build path.

Pika launched Pika Agents, a persona-based interface for making videos, ads, decks and designs by talking to a custom face-and-voice agent. Early creator demos show marketing, research and fashion workflows, but most evidence is still from launch-day tests.

Creator tests showed ChatGPT Images 2.0 making scannable QR codes, color-analysis layouts, study sheets, brand kits, and one-image campaign boards. That pushes the model further into structured graphic work, though typography and brand-rule precision still vary by run.

Glif launched V2 as a chat-based creative agent that chains image, video, voice, and music models, and announced a $17.5 million seed led by a16z and USV. Early demos show multi-model ads and short films being produced inside single conversations instead of manual tool hopping.

Lovart rolled out GPT Image 2 inside its canvas and agents, with creators showing multi-asset campaign generation, live text edits after generation, and a 45% Pro discount. That matters because ad layouts, UI mockups, and marketing sets can stay editable instead of forcing a full re-prompt or a Photoshop pass.

Amir Mushich released Motion Brief, a Claude Project that turns a product shot into motion directions, Seedance prompts and buyer/pricing guidance. Related posts show the same workflow expanding into batch product angles and video demo frames.

Creators say Higgsfield's Marketing Studio can turn one product link into nine ad formats, from UGC to TV spots, with face and brand consistency. Multiple posts also cite about $0.347 per generation, but that pricing detail is user-reported.

Creator and partner posts say OpenArt added Seedance 2.0 with text-plus-reference video workflows, including two-photo animation and AI spokesperson demos. The early material centers on reference-image control rather than low-level model settings, so use it for guided generation.

Runway said one creator finished a short ad in one afternoon, while others published 2-5 minute AI films and shared their stacks. The posts quantified longer production runs, from 398,055 Seedance credits across 113 scenes to multi-tool film pipelines.

Creators on X said recycled-video accounts lost impressions and monetization while original posts surfaced more often. Community reports tie the change to cleaner discovery and better engagement, but the evidence here is still unverified.