# vLLM integrates PyNvVideoCodec to offload video decoding to GPU NVDEC

vLLM says it has integrated PyNvVideoCodec to offload video decoding from CPU to GPU NVDEC, reporting 2x+ throughput at 8×H100. The project says the change ships with CUDA vLLM releases.

Language: en
Time zone: America/Los_Angeles
Current time (UTC, cached as of): 2026-11-10T06:57:00.000Z
Query date (America/Los_Angeles): 2026-11-09

HTML: https://didcodexreset.com/news/007f5aced85e4e7c851f6058.html

Content language: en
Localization state: sameLanguage

Source: vLLM · Published 9/21/2026, 15:34:32

vLLM said it has integrated PyNvVideoCodec to offload video decoding from the CPU to the GPU's NVDEC.

According to the project, the result is **2x+ throughput** at 8×H100, with the CPU bottleneck gone. It said this matters for video captioning at scale (AV training, metadata) and **ships with CUDA vLLM releases**.

Details: [https://vllm.ai/blog/2026-09-18-pynvvideocodec](https://vllm.ai/blog/2026-09-18-pynvvideocodec)

Tags: vLLM, PyNvVideoCodec, NVDEC, Video decoding

[View original post](https://x.com/vllm_project/status/2102142730134814943)