Beatport blocks fully AI tracks, Luma adds editable layers, and cheap 3D assets fail to sell.
Twitch added a control that blocks future use of streams, clips, chat, images, and channel text for Amazon model training, but left it enabled by default. Creators have to opt out under Security and Privacy. The switch does not answer what Amazon may already have used, so this is consent after eligibility rather than consent before collection.
Beatport now rejects music that is wholly or substantially generated by AI and uses Beatdapp detection to identify it across the catalogue. AI-assisted mixing, mastering, stem separation, and audio processing remain allowed when they support human creativity. For creators, disclosure is becoming a distribution requirement that platforms may enforce before a track reaches listeners.
CGTrader says AI-generated assets now make up one in six uploads but only about $1 of every $90 in marketplace revenue. Buyers rank quality above price, and only 5% said purchased AI models worked well. Cheap production created supply, not demand, which is a useful warning for anyone treating output volume as a creative business model.
Heather Cooper used Invideo Agent Two to make the short film DROUGHT, then walked through the workflow and the parts she found useful. The value here is the production breakdown, not another polished demo reel. It gives creators a concrete look at how the agent fits into a short-film process before they commit their own project to it.
This guide structures a faceless script as hook, promise, context, payoff, and close, with word budgets for Shorts and longer videos. It treats the first three seconds as a distribution decision and recommends changing one hook variable at a time. The useful part is the repeatable testing loop, with AI handling volume and human editing restoring specificity.
Chinese microdrama producers are replacing shoots with Seedance 2.0 generations that take days instead of weeks. One studio owner says business is down 70% year over year, while actor Li Dazhi says his pay has halved. The audience appears willing to watch, which makes this a production economics story before it becomes a quality debate.
Luma Layers separates objects, text, and backgrounds so one element can change without regenerating the approved frame. It also decomposes uploaded images, preserves untouched pixels, and supports reusable elements across compositions. For campaign teams, the practical win is cheaper localization and revision work without sending every small copy change through another full generation.
@rourke_heath asks: if agents automate making, what are creatives saving time for? He uses generative tools to execute ideas that might stay in his head. His line is between extending your hands and outsourcing the part that gives the work meaning.