Free downloadable AI video models for 4K workflows: LTX-2.5 and more
TinfoilTricorn · x · 2026-10-12
A curated, October 2026-updated list of open-weight, locally runnable video models for text-to-video, image-to-video and video-to-video generation, editing and enhancement.
- LTX-2.5 (Lightricks): a 22B model family with synchronized audio/video, multi-shot generation, documented 4K finishing pipeline (multistage diffusion + upscaling), ComfyUI and local Python inference. Free commercial use under $10M annual revenue; HF login and license acceptance required.
- Caveats: free weights don't mean free hosted inference, and 4K output may involve generative upscaling.
More from Multimodal
- Mistral showcases vision model for engineering, manufacturing and gigapixel satellite imagery — MistralAI · 2026-10-13
- Mistral says its new model hits 42% on Dense 200 visual grounding, beating frontier models — MistralAI · 2026-10-13
- Pencil-sketch AI image style takes off: neon-lit Tokyo streets drawn in graphite — azed_ai · 2026-10-13
- Gemini-generated pencil-sketch saxophonist nails retro camcorder aesthetic — azed_ai · 2026-10-13
- Struggling to Reliably Change Camera Angles in Qwen Image Edit 2.1 — Numerous_Access_8008 · 2026-10-13
- Codex Drives CapCut Desktop: AI Moves From Generating to Running the Whole Creative Pipeline — HeyAmit_ · 2026-10-13