Shaka Streamer offers a simple config-file based approach to preparing streaming media. It greatly simplifies the process of using FFmpeg and Shaka Packager for both VOD and live content.

Platform support

We support common Linux distributions and macOS.

Windows is not supported at this time due to our use of os.mkfifo, but we are accepting PRs if you'd like to add Windows support. See issue #8.

Getting started

Shaka Streamer requires at a minimum:

• Python 3
• Python "yaml" module
• Shaka Packager
• FFmpeg

See the file PREREQS.md for detailed instructions on installing prerequisites and optional dependencies.

To use Shaka Streamer, you need two YAML config files: one to describe the input, and one to describe the encoding pipeline. Sample configs can be found in the config_files/ folder. Sample inputs referenced there can be downloaded individually over HTTPS or all at once through gsutil:

gsutil -m cp gs://shaka-streamer-assets/sample-inputs/* .

Example command-line for live streaming to Google Cloud Storage:

python3 shaka_streamer.py \
  -i config_files/input_looped_file_config.yaml \
  -p config_files/pipeline_live_config.yaml \
  -c gs://my_gcs_bucket/folder/

Example command-line for live streaming to Amazon S3:

python3 shaka_streamer.py \
  -i config_files/input_looped_file_config.yaml \
  -p config_files/pipeline_live_config.yaml \
  -c s3://my_s3_bucket/folder/

Running tests

We have end-to-end tests that will start streams and check them from a headless browser using Shaka Player. End-to-end tests can be run like so:

python3 run_end_to_end_tests.py

Hardware encoding

For details on hardware encoding support, see the file HARDWARE_ENCODING.md.

Technical details

Shaka Streamer connects FFmpeg and Shaka Packager in a pipeline, such that output from FFmpeg is piped directly into the packager, and packaging and transcoding of all resolutions, bitrates, and languages occur in parallel.

The overall pipeline is composed of several nodes. At a minimum, these are TranscoderNode (which runs FFmpeg) and PackagerNode (which runs Shaka Packager). They communicate via named pipes on Linux and macOS.

If the input type is looped_file, then LoopInputNode is placed before TranscoderNode in the pipeline. LoopInputNode runs another instance of FFmpeg to encode the input file in a never-ending loop, and output to a named pipe. For all other input types, the input files are read directly by TranscoderNode.

If the -c option is given with a Google Cloud Storage URL, then an additional node called CloudNode is added after PackagerNode. It runs a thread which watches the output of the packager and pushes updated files to the cloud.

The pipeline and the nodes in it are constructed by ControllerNode based on your config files. If you want to write your own front-end or interface directly to the pipeline, you can create a ControllerNode and call the start(), stop(), and is_running() methods on it. You can use shaka_streamer.py as an example of how to do this.