🎧 Listen to this post as an audiobook
Audiobook in comic character voices (Qwen3-TTS, local)

A single prompt now returns a ten-second vertical ad. The format everyone wants is UGC: not a polished brand spot, but something that looks like a real customer filmed it on a phone and posted it. So the prompt asks for handheld shake, film grain and a fogged lens instead of asking for quality. You are paying, line by line, to look unpolished. The only part still unaccounted for is the satisfied customer standing in the shot.

User-Generated, Minus the Users

Source: RT @Soaima_Ai: GPT image 2 · twitter

▶ Animated edition — the characters speak for themselves (Korean audio)

Download video

What this means for ThakiCloud

Generating the creative this way is genuinely convenient. The catch is where it renders. Ship the job to someone else’s cloud and your unreleased product shots and brand assets go with it. Paxis hands the loop of writing prompts, generating a hundred takes and culling the usable ones to agents, and Metis serves the image and video models on GPUs we own. A hundred takes, none of which leave the building. You only get one real customer anyway, so the other ninety-nine may as well live in our racks.


An auto-generated comic riffing on this week’s industry news.

Tags: ai-video, on-prem, prompt, sovereignty, ugc-ad

Categories:

Updated: