Steel Ray is Korea's most-watched AI feature film
The 90-minute sci-fi feature was assembled from about 13,000 AI clips on a budget under 1 billion won, and its ticket sales show how hard AI films are to sell.
Launches, updates, demos and everything that shipped.
The 90-minute sci-fi feature was assembled from about 13,000 AI clips on a budget under 1 billion won, and its ticket sales show how hard AI films are to sell.
The psychological thriller-horror was cleared uncut as a short film and used AI for its screenplay, character design, animation, score and visual effects.
Higgsfield demands Comfy stop using its registered mark and its chief executive's face; Comfy says it consulted lawyers and the challenge stands. Everything you need on the fight: the letter itself read page by page, who the two companies actually are, what the outlets have reported, the legal position, the builders' mood, and the 6 November date both sides are now looking at. Updated every three days.
73 percent of Gen Z adults call AI-made people and voices in ads unacceptable, per Bentley-Gallup, while comments and consumer testing split.
P-Video-2-Pro is built on MiniMax H3's open weights, and H3 itself rose to third on Artificial Analysis's text-to-video board.
Promotional rates on fal's MiniMax H3 Max Turbo, Higgsfield's Wan 3.0 API and HeyGen Video close with October, and a two-minute 1080p clip climbs from about $6.25 toward $9.60.
Monthly Pro and Plus cover only the two image models, while annual plans add Seedance 2.5 at up to 20 seconds.
The clip rode the real October sinking of the GFS Pride, but detectors found a SynthID watermark and visual glitches.
The image-to-video model is built on Alibaba's Wan 2.2, and TBC says lessons from lab-grown neurons make it about five times faster.
The enterprise platform, used by 65% of the Fortune 500 according to Vyond, requires an additional subscription and a custom-quote price.
Founder Luke Igel says Kino stores, cuts and reviews footage in one place, with editing decisions living as code on a timeline that people can override.
The KAIST paper's latent 3D Gaussian memory tops DL3DV baselines and uses 266 to 417 times fewer 3D primitives than NeoVerse, but no weights are out yet.
The model translates video, clones each speaker's voice and lip-syncs the result, while a Pro tier keeps background audio and costs 18 cents a second.
Playable Games runs on every Higgsfield plan, but turning a game into an interactive ad for nine networks takes an Enterprise plan.
Pika's post-trained MiniMax H3 turns out a 15-second 768p clip in as little as seven seconds of denoising time, and the API starts at a cent a second.
The two hosted endpoints rebuild a reference clip of up to 15 seconds as one HTML page or a looping GLSL shader, priced from $0.20 a second.
The Apache 2.0 model writes, renders and reviews its own frames inside Claude Code, for motion graphics and explainers rather than photoreal footage.
Nine general coding agents each got one hour to build a finished video with sound using only code, and one of them delivered as few as 179 of 239 videos.
MotionVFX's tools now arrive inside a $12.99-a-month Creator Studio subscription, while the one-time version of Final Cut Pro costs $299.99.
Maya Research published Apache 2.0 weights and inference code on October 25, with 20 voices and cloning from 3 to 10 seconds of speech.