SVT Encore¶

Encore

Scalable video transcoding as a service, built on FFmpeg and Spring Boot.

Encore is an open-source transcoding tool created by SVT (Sveriges Television) to scale and abstract the transcoding power of FFmpeg, offering a simple REST API for transcoding — Transcoding-as-a-Service. Encore has been in production at SVT since 2019 and was open sourced in 2021.

Encore is aimed at the advanced technical user that needs a scalable video transcoding tool — for example, as a part of a VOD (Video On Demand) transcoding pipeline.

Key features¶

Scalable — separate web API and worker components, Redis-backed job queues, configurable concurrency
Profile-driven transcoding — YAML-based transcoding profiles with support for dynamic parameters via SpEL expressions
Codec support — H.264 (x264), H.265 (x265), any FFmpeg-supported codec via generic video encode, plus audio codecs (AAC, AC-3, etc.)
Thumbnail generation — single thumbnails and thumbnail map sprite sheets
Quality metrics — VMAF quality measurement
Progress callbacks — HTTP callbacks for job progress monitoring
Segmented transcoding — parallel chunk-based transcoding for faster processing
Security — optional basic authentication with user and admin roles
OpenAPI — built-in Swagger UI for API exploration

Design philosophy¶

Encore is designed as a standalone component intended to operate within a larger automated system. Rather than being an all-in-one transcoding suite, Encore focuses on doing one thing well — executing transcoding jobs reliably and at scale. These principles guide its design:

Automation-first — Encore is built primarily to work inside automated workflows and does not feature a UI. Job feedback is handled through progress callbacks, making it straightforward to integrate into end-to-end transcoding pipelines.
Priority queues — Jobs are managed through priority queues, allowing urgent work to be processed ahead of lower-priority tasks.
Minimal dependencies — Redis is the only external dependency beyond FFmpeg. Using a single data store for job storage, queuing, and pub/sub keeps deployment simple and makes it easy to scale horizontally.
FFmpeg command-line builder — Rather than reimplementing FFmpeg's vast parameter space, Encore constructs FFmpeg command lines directly. This gives full access to FFmpeg's flexibility without Encore needing to wrap every option.

Architecture¶

Encore consists of two services that communicate via Redis:

graph LR
    Client -->|REST API| Web[encore-web]
    Web -->|enqueue job| Redis[(Redis)]
    Redis -->|poll job| Worker[encore-worker]
    Worker -->|transcode| FFmpeg[FFmpeg]
    FFmpeg -->|output files| Storage[File Storage]
    Worker -->|update status| Redis
    Client -->|poll status| Web
    Worker -.->|progress callback| Client

encore-web — the REST API service. Accepts transcoding jobs, serves the Swagger UI, and optionally processes jobs itself.
encore-worker — a headless worker that polls Redis for jobs and runs FFmpeg to transcode them. Scale horizontally by running multiple workers.
Redis — stores job data and acts as the job queue. Used for pub/sub (cancellation, segmented transcoding progress).
FFmpeg — the transcoding engine. Must be installed on all machines running encore-web (if polling is enabled) or encore-worker.

What Encore is not¶

Not a live/stream transcoder
Not a video packager (use a tool like Shaka Packager for that)
Not a GUI application

Built with¶

Kotlin
Spring Boot
FFmpeg
Redis (via Lettuce)
GraalVM (native image compilation)

License¶

Encore is licensed under the EUPL-1.2-or-later license.