Local video generation on 64GB RAM: ComfyUI + LTX-2.5 struggles with character consistency
krustyy · reddit · 2026-10-22
A self-hosting beginner is trying to generate a 5-minute video of Slimer (Ghostbusters) on a Ryzen AI Max 395+ with 64GB RAM, using ComfyUI locally, for a Pepper's Ghost "containment unit" display.
Current setup: a Gemini-generated first frame, then LTX-2.5 producing 720p/24fps/10-second clips (10 min render each). After 10 generations, every clip breaks character consistency — legs appear, a human voice slips in, the body shape and color keep changing — despite heavy prompt tweaking. He needs 300 consistent clips to stitch together, then plans to upscale to 4K. The post includes the full prompt and asks for model recommendations that fit his hardware.
More from Multimodal
- CrossFit-GS uses out-of-fold cross-validation to fix pose drift in COLMAP-free 3D Gaussian Splatting — kwangmoo_yi · 2026-10-22
- Hermes agent now edits videos directly in DaVinci Resolve from a script — alexcovo_eth · 2026-10-22
- Thrixel adds auto-rigging for characters, creatures, and robots with Mixamo/Unity/Roblox support — RanaHanocka · 2026-10-22
- 360-orbit workflow achieves perfect face consistency across angles from a single frontal photo — Future_Challenge_910 · 2026-10-22
- Image-to-video in ComfyUI without a local GPU: Forma custom nodes, pay-as-you-go from $5 — no3us · 2026-10-22
- Six image models compared on a tricky two-input edit: style swap, text removal, and face replacement — ZootAllures9111 · 2026-10-22