code-dot-org / p5-replay

Service to generate mp4 files from JSON animation specs from Code.org projects

p5-replay

Generates a movie file based on a p5 animation replay file, and uploads it to S3.

Based on serverless-ffmpeg

Building/deploying

• Make sure the Docker daemon is installed and running (https://docs.docker.com/install/)
• The bucket name specified in template.yml must be unique globally because of S3's global bucket namespace - you might need to change it if it already exists (an error will occur when deploying if so)
• [TODO] Don't change the node-canvas version away from 2.0.0-alpha.13 - later releases seem to cause node-gyp problems
• Run the build.sh script, which will create a Docker image to build the application binaries
  • The first time this runs the Docker part will take a while to download, and there may be some red error text, but that's fine, it should complete quickly the next times it is run.
• Run the deploy.sh script to deploy the application CloudFormation stack.

Local Testing

Local testing is possible via sam local and Docker:

• Install the AWS SAM CLI
• yarn test

Remote Testing

The Serverless deploy should display two endpoints after completion:

• GET - https://[UNIQUE_ID].execute-api.us-east-1.amazonaws.com/dev/runTest
• POST - https://[UNIQUE_ID].execute-api.us-east-1.amazonaws.com/dev/render

The runTest endpoint will generate a video based on the replay.json file deployed; render is a POST method that uses the body of the request. You can use the replay.json contents as the content of the POST request, just make sure the content type is application/json (use raw mode if you're sending requests via Postman).

Sample Data

node ./scripts/generateData.js number_of_frames number_of_sprites > ./test/fixtures/replay.json will generate some random sample animation data with a given frame and sprite count.

Built With

• FFMPEG
• node-canvas
• Docker