Issue #7 · Wednesday, July 8, 2026 · 7 stories

📸 Now Anyone Can Put Your Instagram Face in an AI Image 🎭 An AI Actress Lands a Real Movie

Plus Conair's AI ad cut cost-per-view 14%, and a director's on-set AI VFX pipeline that adds crew.

This week the ownership question got real: Meta's Muse model now lets anyone tag a public Instagram account and generate AI images of that person, opt-out only, with no heads-up. An AI 'actress' landed a feature film while her studio will not say what she was trained on. On the tools side, Comfy MCP lets an agent run your whole ComfyUI pipeline from chat, NotebookLM turns your documents into vertical shorts, and a new LoRA family keeps characters consistent across shots.

NEWS

Meta Lets Anyone Generate AI Images of You From Instagram Unless You Opt Out · 3 min read

Meta's new Muse Image model lets anyone tag a public Instagram account in a prompt to generate an AI image using that user's likeness. Public profiles are automatically opted in, and users must manually disable this in settings. No notification is sent when your content is used, and existing AI images remain after opting out. For creators who share content publicly, this means their likeness can be remixed without consent unless they take action.

AI Actress Sparks Union Backlash Over Secret Training Data · 3 min read

Particle6's AI actress Tilly Norwood will star in the film 'Misaligned' but the studio won't disclose what data was used to train her. SAG-AFTRA and actors are pushing back over concerns that performers' work was used without permission. The studio claims a hybrid workflow with 30 human creatives and 2,000 iterations, but the transparency gap is the real story for anyone building AI video tools.

Conair's AI Video Ad Cut Production Time From Months to Four Weeks · 5 min read

Conair tested Amazon's Creative Agent AI to produce a 15-second Cuisinart food processor video. The AI handled 80% of the concept, resulting in 18% higher detailed page views and 14% lower cost per detail page view compared to a traditional video. Production time dropped from months to four weeks, though a human agency still had to fix logos and colors. Conair plans to scale the tool but says humans remain necessary for brand consistency.

ANALYSIS

Filmmaker Builds AI VFX Pipeline That Augments, Not Replaces, Real Footage · 7 min read

Director Jon Erwin details a practical AI workflow for 'Young Washington' and 'Moses' that synthesizes owned assets (location footage, actor consent, wardrobe) into photorealistic VFX in real time on set using Unreal Engine and gaussian splatting. The approach brings VFX into production rather than post, reduces costs, and enables safer location shooting. For builders, it's a case study in using AI to augment existing assets and workflows instead of generating from scratch.

TOOLS

AI Agents Drive ComfyUI Image/Video Generation via MCP · 3 min read

ComfyUI is the free, node-based engine most of the open-source AI video and image scene runs on. Comfy MCP lets an AI agent like Claude run those workflows from chat: connect it once, then ask the agent to generate images, video, audio or 3D and pull the files back. It is in open beta, so you can automate ComfyUI without wiring nodes by hand or babysitting a local GPU.

NotebookLM Now Turns Your Documents Into Vertical Short Videos · 1 min read

NotebookLM, Google's research-notebook tool, can now turn the documents you upload into short vertical videos, the kind that fit TikTok, Reels and Shorts. It is free, works on mobile, and is built for repackaging something you have already written into short-form video without editing it yourself. It is a fast way to spin up an explainer clip, though you give up fine control over how the finished video looks.

RefControl LoRAs Add Depth, Pose and Canny Control to FLUX.2 Klein · 1 min read

FLUX.2 Klein is an open image model that struggles to keep a character consistent from one shot to the next. RefControl is a family of free add-on LoRAs that lock a reference image to a control map (depth, pose, outline) so the face and style hold while you change the pose. That is the consistency you need for video reference frames. It is community-made, so quality varies by control type.