This page provides a technical overview of the Transcoder API, including features, terminology, and useful concepts. The Transcoder API implements a REST and RPC API that lets you submit, monitor, and manage transcoding jobs in Google Cloud. To submit jobs using the Transcoder API, you first upload media assets to Cloud Storage. After processing a job, the Transcoder API saves the resulting media back to Cloud Storage.
The Transcoder API includes support for the following features:
- Output in different container formats, including MPEG-4 (MP4), Dynamic Adaptive Streaming over HTTP (DASH, also known as MPEG-DASH), and HTTP Live Streaming (HLS)
- Output at different bitrates and resolutions
- Enhance the video output programmatically, including:
- Configure low-level encoding parameters, such as the bitrate
- Remix existing media content using comprehensive edit lists
- Specify audio loudness normalization
Supported inputs and outputs
The following table shows the supported input and output formats for the Transcoder API.
|Supported input and output formats|
|Containers||AVI, GXF, MKV (Matroska), MOV, MPEG2-TS, MP4, MXF (OP1A only), WebM, WMV|
|Video Codecs||DNxHD, DV/DVCProHD, DV25, DV50, H.261, H.262, H.263, H.264 (AVC), H.265 (HEVC), MPEG-1, MPEG-2, MPEG-4 part 2, ProRes, Theora, VC-1, VP8, VP9, XDCAM|
|Color Space||rec.601, rec.709, and rec.2020|
|Audio Codecs||AAC, AC3, AIFF, E-AC3, MP3, Opus, PCM, WAV, WMA, WMA2, Vorbis|
|Captions||SCC, SRT, VTT|
|Containers||Apple HLS (with MPEG2-TS, fMP4), MPEG-DASH (with fMP4), standalone MP4|
|Video Codecs||H.264 (AVC), H.265 (HEVC), VP9|
|Color Space||rec.601 only; the input color space is not converted so a color change may be observed|
|Audio Codecs||AAC-HE, AAC-HEv2, AAC-LC, AC3, E-AC3, MP3|
|Images||JPG tiles, single images|
The following table shows limits per transcoding job.
|Total input size||400 GB|
|Total output size||400 GB|
|Maximum ||24 hours|
|Maximum number of video streams||30|
|Maximum number of audio streams||50|
|Maximum number of text streams||50|
|Maximum number of mux streams||100|
|Maximum number of manifests||100|
|Maximum number of spritesheets||10|
This section provides important concepts regarding video files and how they are used with the Transcoder API.
Components of a video file
Each video file has a container, which is the wrapper for the entire file. The Transcoder API uses MuxStreams to define the container. Each container includes a set of ElementaryStreams to define the encoding of the video, audio, and caption text tracks for the file. Video and audio are compressed using codecs.
In the following example, the video is compressed using H.264 and the audio is compressed using AAC. Both are placed in an MP4 container.
Figure 1. Components of a video file (left) and an example MP4 file (right).
Streaming protocol structure
For streaming media, content providers encode the same content at multiple bitrates (measured in Kilobits per second). The provider then generates a manifest file that points to the different files with different bitrates. The streaming media player uses the manifest file to select the appropriate file, and then pulls video a few seconds at a time. Each bitrate may be a single file or multiple short files, depending upon what the player supports.
Different resolutions of media are typically encoded to target different bitrates. For example, lower bitrates are encoded at standard definition (SD) instead of high definition (HD). The set of bitrates, resolutions, and codecs is called the Adaptive Bitrate (ABR) ladder. Streaming content providers may tune their own ladder based on CDN costs, user device types, bandwidth in-region, and other factors.
Figure 2. Components of a streaming protocol structure (top) and an example HTTP Live Streaming (HLS) implementation (bottom).
This section provides a glossary of useful terms for working with the Transcoder API.
An ad break is a short advertisement that appears before or during media playback. The Transcoder API supports an ad break keyframe in the job configuration. The Transcoder API does not insert or play ads, or stop media playback; video player clients are responsible for handling the keyframe.
An atom is a foundational data structure for defining a video's metadata and location. A video can comprise a complex nested hierarchy of different types of atoms, including audio, edit, and text atoms.
An audio atom maps the audio from an elementary stream to an edit list.
Codec type and profile
When selecting the codec for a video stream, you specify the codec type,
such as H.264 and the profile, such as the default
A container is a wrapper that describes the relationship between the various components in a multiplexed stream, including media files and metadata. The Transcoder API supports the MP4, MPEG-DASH, and HLS container formats.
An edit atom defines the start and end offsets for the individual segments of a stream that you want to combine in an edit list.
An edit list defines a sequence of edits as a timeline for the resulting file or manifest from a transcoding job.
An elementary stream is an encoding of an input file, such as an audio, video, or caption text track. You must package elementary streams before mapping and sharing the stream to different output formats.
Entropy encoding is a form of lossless compression that the Transcoder API supports. When configuring jobs, you can specify either the Context-Adaptive Variable-Length Coding (CAVLC) or Context-Adaptive Binary Arithmetic Coding (CABAC) entropy coders.
A job is the basic unit for managing work with the Transcoder API. When you submit a job to the Transcoder API, it is processed asynchronously in a geographic location known as a region. You can list and manage all jobs for a region.
The lifecycle of a job includes three steps: prepare, transcode, and package.
- Download inputs from Cloud Storage
- Analyze inputs
- Validate inputs
- Run transcoding operations on inputs
- Stitch inputs
- Multiplex inputs
- Upload outputs to Cloud Storage
A job configuration represents many of the various settings you can customize when creating and submitting a job to the Transcoder API. You can specify configuration settings such as edit lists and where to insert ad break tags in an output manifest. You can create reusable job configurations as job templates for use in a Google Cloud region.
By default, the Transcoder API applies a preset template called
preset/web-hd for populating a job configuration. This job configuration
produces the following output files:
manifest.m3u8: The primary playlist for an HLS media stream. This file contains references to playlists for the high definition (HD) variant of the output and the standard definition (SD) variant of the output.
media-hd.m3u8: Playlist for the high definition variant
media-hd0000000000.ts: High definition video segment file
media-sd.m3u8: Playlist for the standard definition variant
media-sd0000000000.ts: Standard definition video segment file
manifest.mpd: The playlist for an MPEG-DASH media stream. This file contains references to video-only and audio-only segment files.
audio-only0000000000.m4s: Audio-only segment file
video-only-hd0000000000.m4s: High definition video-only segment file
video-only-sd0000000000.m4s: Standard definition video-only segment file
sd.mp4: Standalone standard definition video file
hd.mp4: Standalone high definition video file
You can create and manage your own custom job templates and specify them when creating jobs.
Preprocessing is the stage of a job that takes place prior to the main transcoding. When configuring the preprocessing stage, you can apply filters such as color, denoising, and deblocking. Deblocking can help reduce block artifacts as a result of compression.
A preprocessing configuration represents the settings for filters you can apply to a video before the transcoding stage of a job. You can apply a variety of filters including color saturation, deblocking, and denoising.
Rate control mode
The rate control mode indicates whether to process a job using either the constant rate factor (CRF) or variable bitrate (VBR) modes. CRF ensures constant quality throughout the processed media assets. VBR optimizes the encoding process to reduce the file size of the processed media assets. For streaming use cases, select the VBR rate control mode. For archival purposes, select the CRF rate control mode.
A manifest is a description of the available contents and metadata of an adaptive media stream to a client. In the Transcoder API, you can configure a job to output a manifest with a file name, a list of multiplexed streams, as well as the type of the manifest, which must be either HLS or MPEG-DASH.
A text atom maps the text from an elementary stream to an edit list.
A text stream encodes text data associated with a video, such as closed captions or subtitles.