Skip to content

SYSTEM Cited by 1 source

Meta Scalable Video Processor (MSVP)

Definition

MSVP (Meta Scalable Video Processor) is Meta's custom ASIC for video transcoding, deployed only inside Meta's infrastructure. MSVP's role in Meta's pipeline is the same as NVIDIA's NVDEC/NVENC or Intel's Quick Sync Video in other pipelines: hardware-accelerated encode/decode off the general-purpose CPU (Source: sources/2026-03-09-meta-ffmpeg-at-meta-media-processing-at-scale).

The canonical Meta MSVP post is meta-scalable-video-processor-MSVP, referenced from the 2026-03-09 FFmpeg post; it is the source-of-truth for MSVP's internal architecture and is not yet ingested into this wiki.

Integration with FFmpeg

MSVP plugs into FFmpeg through the same standard hardware-acceleration APIs that already support NVDEC/NVENC, AMD UVD, and Intel QSV โ€” see concepts/hardware-accelerated-video-codec-api. Per the post: "enabling the use of common tooling across different hardware platforms with minimal platform-specific quirks."

Choosing to integrate through the existing hardware-accel abstraction (rather than adding an MSVP-specific CLI surface) is what lets Meta's video transcoding pipelines move between MSVP-backed and software-backed execution without pipeline rewrites.

Why patches stay internal

MSVP hardware lives only inside Meta. External FFmpeg developers have no way to test or validate MSVP-specific code paths. The post is explicit that this is the correct reason to not upstream these patches โ€” canonical wiki instance of patterns/keep-infrastructure-specific-patches-internal:

"As MSVP is only used within Meta's own infrastructure, it would create a challenge for FFmpeg developers to support it without access to the hardware for testing and validation. In this case, it makes sense to keep patches like this internal since they wouldn't provide benefit externally."

The accepted operational cost is the reverse-rebase burden: "We've taken on the responsibility of rebasing our internal patches onto more recent FFmpeg versions over time, utilizing extensive validation to ensure robustness and correctness during upgrades."

Seen in

Caveats

  • The FFmpeg post does not disclose MSVP's chip specifications, deployment fleet size, codec support matrix, or throughput numbers. All of that lives in the separate 2023 MSVP post.
Last updated ยท 319 distilled / 1,201 read