Local LTX 2.5 Generation: 12 Minutes for 20 Seconds of Video?
huynguyend · reddit · 2026-08-30
A user reports that generating 20 seconds of video using the LTX 2.5 Distilled INT 8 model takes 12 minutes on a Lenovo P16 Gen 2 with an NVIDIA RTX 3500 Ada (12GB VRAM) and 64GB RAM. The workflow involves image and audio inputs. The user is seeking confirmation if this latency is normal and asking for potential optimization methods, highlighting the performance challenges of running video models on consumer hardware.
More from Infra
- Data center boom drives massive demand for blue-collar jobs and freight — surmenok · 2026-08-30
- workweave/router: open-source model router cuts agentic costs 40-70% in <50ms — workweave · 2026-08-30
- ODS turns any PC into a local AI server: inference, RAG, voice, agents — Osmantic · 2026-08-30
- Latency Patterns for Faster AI Apps: Beyond Model Inference — bibryam · 2026-08-30
- Tenstorrent benchmarks for Qwen3.7-27B reveal performance on dual p300c setup — DustNearby2848 · 2026-08-30
- Bot Mesh: A social network with identity and payments for AI agents — Daniel_Farinax · 2026-08-30