Blender on Lambda

Render Blender scenes on the distributed serverless cloud!

This repository is based on the architecture laid out by TAS-PC.

This repository aids in the following tasks:

• Setting up cloud infrastructure in AWS for rendering blender scenes
• Executing multiple concurrent jobs for rendering each frame
• Uploading and downloading scene and support files
• Removing infrastructure

A few notes on the chosen architecture

The infrastructure created is called Serverless, through an AWS service called Lambda functions. These are charged per-use, and have a monthly free tier that costs nothing to use. AWS Lambdas do not currently support GPU rendering, but are easily scaled to concurrent executions, saving time.

How to use

1. Create an AWS Account.
2. Download the AWS CLI. Install it and configure your credentials (aws configure).
3. Download terraform.
4. Clone this repository
5. Create a blender-lambda.tfvars file for configuration. Use these values as startup, but modify them as you will.

   default_tags = {
       "project" = "blender-lambda"
   }
   
   producer_api_gateway_name = "blender-lambda-api"
   
   producer_lambda_source_path = "./blender-lambda-producer"
   
   producer_invocation_route_key = "POST /render-job"
   
   producer_lambda_function_name = "blender-lambda-producer"
   
   producer_ecr_repo = "blender-lambda-producer"
   
   consumer_lambda_source_path = "./blender-lambda-consumer"
   
   consumer_lambda_function_name = "blender-lambda-consumer"
   
   consumer_ecr_repo = "blender-lambda-consumer"
   
   queue_name = "blender-lambda-queue"

6. Execute ./apply.sh and accept the changes to have the infrastructure be created on your AWS account.
7. Execute ./upload_files.sh <sceneFile> <file1> <file2> <file3> to upload your files (the scene file and any supporting meadia files).
8. Execute ./render.sh <sceneFile> <file1> <file2>... to start the render job.
9. Wait.
10. Execute ./download_rendered.sh to download the results of the render.
11. Execute ./destroy.sh to remove all the infrastructure from your AWS account.

More information

This work is based on these previous investigations:

• TAS-PC
• Rendering Blender Scenes in the cloud with AWS Lambda, by JR Beaudoin
• Blender-docker by the NYTimes Research Team