Question: Can llama.cpp handle mp4 video inputs like vLLM?

trashacct383 · reddit · 2026-08-29

User asks if llama.cpp supports direct mp4 video file input (e.g., for Qwen3.8-27B models). While vLLM allows sending mp4 directly to the API with sampling parameters, llama.cpp seems to require a separate sampling pipeline and feeding still frames. The user seeks a more direct solution.

Original post →

More from Infra

Infra channel →