Encode a movie from another Deadline Cloud job's output with FFmpeg

The ffmpeg_movie_from_job_output job bundle downloads the rendered output of another completed job in the same queue and encodes the image sequence into an MP4 video using FFmpeg. Use it as a post-processing utility — for example, to automatically assemble frames into a movie after a Blender or Maya render job completes.

For encoding a local image sequence (no source-job download), see Encode video with FFmpeg on Deadline Cloud.

A pre-submission hook (inject_s3_settings.py) runs at submission time on your workstation and looks up the queue's job attachment Amazon Simple Storage Service (Amazon S3) bucket configuration. The hook writes the settings to a JSON file that gets uploaded as a job attachment, so the worker can download files from Amazon S3 without needing Deadline Cloud API permissions on the queue role. On the worker, the job uses the deadline.job_attachments Python API to download the source job's output, then runs FFmpeg to encode the frames into an H.264 MP4 with BT.709 color space metadata.

Bundle hooks are off by default. Enable them once before submitting:

deadline config set settings.allow_bundle_hooks true

The job requires FFmpeg from the conda-forge conda channel. The source job and this job must be in the same queue, because they share the same job attachments Amazon S3 bucket.

Submit with the source job's ID:

deadline bundle submit ffmpeg_movie_from_job_output/ \
  -p SourceJobId=job-id \
  -p FrameRate=30 \
  -p OutputFilename=my_render.mp4

To restrict the download to a single step's output, also pass SourceStepId.