Amazon Elastic Transcoder
A media transcoding service that converts video and audio files stored in S3 into multiple device-compatible formats
Overview
Amazon Elastic Transcoder is a cloud-based transcoding service that converts video and audio files stored in S3 into formats playable on various devices including smartphones, tablets, PCs, and smart TVs. Simply specify an input file and select a preset (output format configuration template) to automatically perform codec conversion to H.264, VP8, VP9, resolution changes, and bitrate adjustments. It manages processing through the pipeline concept, enabling priority-based job queuing. It also supports conversion to HLS (HTTP Live Streaming) format, automating adaptive bitrate streaming delivery preparation. Pricing is pay-per-use based on minutes of transcoded video.
Pipeline and Preset Design
Elastic Transcoder processing is structured around three concepts: pipelines, jobs, and presets. A pipeline is a processing channel linking input and output S3 buckets, and multiple pipelines can be created for different purposes (e.g., high-priority production, low-priority batch). A preset is a template defining detailed output format settings including codec, resolution, bitrate, frame rate, and audio settings. AWS-provided system presets (Generic 1080p, iPhone 4S, Web, etc.) let you immediately start using settings for major devices. Custom presets offer fine-grained control over bitrate upper/lower limits, keyframe intervals, and profile levels (H.264 Baseline/Main/High), optimizing the balance between quality and file size. Thumbnail generation can be configured simultaneously, extracting frames at specific points in the video as JPEG/PNG for use as preview images.
HLS Conversion and Building Adaptive Bitrate Delivery
One of Elastic Transcoder's key features is conversion to HLS (HTTP Live Streaming) format. HLS splits video into short segments (typically 10 seconds) managed by a manifest file (.m3u8), playable on a wide range of devices including iOS, Android, and web browsers. To achieve adaptive bitrate streaming, generate HLS outputs from the same input video at multiple bitrates (e.g., 400kbps, 800kbps, 1500kbps, 3000kbps) and bundle them with a master playlist. The player automatically selects the optimal bitrate based on the viewer's network bandwidth, minimizing buffering while delivering the highest quality. A common design combines this with CloudFront for delivery, caching HLS segments output to S3 via CloudFront. For encryption, AES-128 encryption is supported, and combining it with a separate key management server prevents unauthorized content downloads.
Job Management and Event-Driven Architecture Design
Elastic Transcoder jobs execute asynchronously, with completion notifications received via SNS topics as the standard design. Associating an SNS topic with a pipeline triggers notifications for job start, completion, error, and warning events. In an event-driven architecture, an effective design triggers Lambda on video upload to S3, automatically creating Elastic Transcoder jobs. Incorporating logic within Lambda to retrieve input video metadata (resolution, duration) and dynamically select appropriate presets enables automatic handling of various input formats. After job completion, another Lambda receives the SNS notification, stores transcoding result metadata in DynamoDB, and returns playable URLs to the frontend, building a complete pipeline. For error handling, a robust design submits failed transcodes to a retry queue (SQS), moving to a dead-letter queue for manual review after a set number of retries. From a cost optimization perspective, since pricing is based on conversion time (per minute), designing to generate only the resolutions needed for actually-viewed devices while omitting unnecessary high-resolution outputs is important.