What LTX 2.3 Can Do - Part 1: Faces & Emotion | Free AI Video, Local on a Mac

What LTX 2.3 Can Do — Part 1: Faces & Emotion (Free AI Video, Local on a Mac) Every face in this film was generated by LTX 2.3, a free, open-weights AI video model — running entirely locally on an Apple Silicon Mac. No cloud, no NVIDIA, no subscription for the model itself. HOW IT'S MADE — the setup: • LTX 2.3 (open-weights, runs locally, self-built q8 version) — the free video model • An Apple Silicon Mac — no NVIDIA GPU needed. About 8–11 minutes to render each shot • Nano Banana / Higgsfield — to generate the still "start image" for every shot • ElevenLabs — for the narrator voice and the music • Claude Code — the AI agent that directs the whole pipeline: it writes the prompts, runs the renders and edits the film • The method is image-to-video: one still image plus one short motion prompt per shot, a slow camera, and one clear movement. Close-up faces are where the model is strongest. Total ongoing cost: only ElevenLabs (~€20 / month). The video model itself is free. HONEST NOTE — this isn't magic. Small things still go wrong: a movement running the wrong way, a detail landing in the wrong place, water on a face instead of on the glass. Almost all of it can be cleaned up with better prompting and a little iteration — and a lot more is still possible than what you see here. That's the whole point of this channel: to show honestly what works, what doesn't, and how far a free local model already goes. ▶ Part 2 — Worlds, Nature & Creatures — is on the channel. #LTX2 #LTXVideo #AIVideo #FreeAI #LocalAI #ClaudeCode #OpenWeights #MadeOnAMac