← Back to archive
Issue #21 · Monday, August 10, 2026 · 7 stories

🐶 Warner Bros AI Ad Torn Apart 🎤 Nobody Can Prove Rubberz Is AI ⏳ Gemini Omni Free

Ten free Gemini Omni videos before the offer closes, plus what the Fenix Flexin fight is about.

Warner Bros. put an AI dog ad on Instagram this week and the comments took it apart within hours. Fenix Flexin's 'Rubberz' stands accused of being entirely machine-made, and releasing the Pro Tools files has not settled it either way. Put those two together and the rule is uncomfortable: audiences do not punish AI use, they punish AI they can see. Everything else in this issue quietly lowers that visibility, from FLUX 3 Video's 20-second clips with lip-synced dialogue in 14 languages, to a ComfyUI node that stitches long MiniMax H3 shots without the character changing face, to ElevenLabs dubbing that holds the original performance across 90 languages. Worth deciding where your own line sits before the tools make it invisible for you.
NEWS

AI Dog Ad for Sci-Fi Film Gets Blasted on Instagram · 2 min read

Warner Bros. Discovery released an AI-generated promotional video for its sci-fi thriller 'The End of Oak Street' in partnership with dog app DogPack. The ad featured AI-animated dogs from a podcast interviewing a character's dog. Instagram comments called it a 'huge f**k up' and said it killed interest in the film. Facebook reactions were more positive. The backlash shows the reputational risk studios face when using AI in marketing, as public sentiment remains broadly hostile toward AI-generated visuals in entertainment.

Google Extends Free Gemini Omni Video Generation to August 11 · 3 min read

Google extended free access to Gemini Omni through August 11, letting users generate up to ten AI videos without a subscription. The tool accepts text, image, and video inputs for creating, editing, and remixing clips. Most advanced video tools require payment, so this free tier gives builders a low-risk way to evaluate Google's video generation model.

FLUX 3 Video opens up: 20 seconds, native audio, lip sync in 14 languages · 1 min read

Black Forest Labs has made FLUX 3 Video generally available through its own API and select partners. This is the video half of the FLUX 3 model that landed in July, not a new model: clips up to 20 seconds at HD or 1080p, with native audio and lip-synced dialogue across English dialects, Chinese, Spanish, French, German, Japanese, Portuguese, Russian, Italian, Indonesian, Turkish, Hindi and Punjabi. BFL puts it at an Elo of 1135 on text-to-video and says it ties Seedance 2.0 while beating everything else on image-to-video, though those are internal evaluations and not a third-party board. The announcement carries no pricing, so budget on it at your own risk until BFL publishes rates.

AI Generates Hitchcock's Vertigo Zoom from a Single Still · 5 min read

The Honor 600 Pro uses AI Image-to-Video 2.0 to create the iconic dolly zoom effect from one still photo. The same system also produces bullet-time and drone camera movements. Output clips are short and characters can be inconsistent. It's a consumer novelty, not a replacement for a film crew.

ANALYSIS

Viral Hip Hop Track 'Rubberz' Under Fire for Undisclosed AI Use · 3 min read

Fenix Flexin's viral song 'Rubberz' is accused of being entirely AI generated using the Trebolo/Sonauto platform. The artist denies the claims and released Pro Tools files, but skepticism persists. The debate highlights the gap between public enjoyment and the ethics of undisclosed AI in commercial music, a tension that affects builders creating AI content tools or detection systems.

TOOLS

Stitching long MiniMax H3 clips without losing motion, audio, or your character · 1 min read

The open-source scene has converged on a fix for H3's biggest practical limit: clips that fall apart the moment you join them. This ComfyUI node carries latent motion, frames, and audio context from one output into the next, and takes extra character and scene references so a face survives across a joined sequence. It now works in Ref2V, and adds the unglamorous production plumbing that matters on a long render: atomic checkpoints, partial assembly, and resume after a crash. Requires ffmpeg, GPL-3.0, roughly 30 stars and 28 commits, forked from NikoDemon80's original with attribution intact. Small project, real engineering, and the difference between a 5-second demo and something you can cut into an edit.

Dubbing API Preserves Tone and Emotion Across 90+ Languages · 1 min read

ElevenLabs launched Dubbing v2 via its ElevenAPI, preserving original performance tone and emotion across 90+ languages with sync aware translation. The update improves regional accent accuracy and handles background audio, music, and multiple speakers. The full pipeline from translation to dubbing runs through a single API call, with enterprise options for granular editing.