Job bundles are the easiest way to define your jobs for AWS Deadline Cloud. They encapsulate an Open Job Description job template into a directory with additional information such as the files and directories that your jobs need. Read more about how to build a job bundle in the Deadline Cloud developer guide. See the example Blender job submission below for more about submitting these jobs to your farm.
This table covers every immediate user-selectable sample directory or collection in job_bundles/.
Nested collections provide their own complete indexes where applicable.
| Sample | What it demonstrates | Start here when |
|---|---|---|
| 3ds Max V-Ray denoiser | V-Ray rendering, frame chunking, and VRIMG-to-EXR conversion with denoising data | You render 3ds Max scenes with V-Ray and need denoised EXR output |
| After Effects one-task render | Running an entire aerender frame range as one task |
A composition must stay on one worker for the full render |
| Arnold standalone render | Rendering Arnold .ass files with kick |
Your scenes are already exported and do not need Maya at render time |
| CARLA autonomous-driving simulation | GPU container simulation, parameter sweeps, and multi-sensor capture | You want to distribute autonomous-driving scenarios |
| Blender render | A compact frame-parallel DCC render job with application packages | You need a minimal Blender or general CLI-render starting point |
| Blender turntable to Flow | Rendering, encoding, thumbnail extraction, and publishing to Flow Production Tracking | You are building a render-to-review studio workflow |
| Blender wedge render from CSV | A pre-submission hook that expands CSV rows into task parameters | A spreadsheet defines your wedge variations, shot list, or parameter sweep |
| Bash CLI job | Submitting a multi-line shell script with an attached data directory | You want to run an ad hoc command-line workload |
| Copy S3 prefix to job attachments | Distributed hashing and copying into content-addressable job attachment storage | Existing S3 datasets should be staged without workstation uploads |
| Custom submitters | A collection of in-application submission interfaces, including Maya | Artist context and DCC state require a custom submission UI |
| ESMFold prediction | Parallel protein structure prediction, validation, and rendering | You need a GPU bioinformatics pipeline from FASTA to PDB outputs |
| FFmpeg encode video | Encoding a numbered image sequence into MP4 | You need a standalone render-output encoding utility |
| FFmpeg movie from job output | Downloading another job's output and encoding it downstream | Post-processing should be a separately submitted follow-up job |
| FLUX.2 Klein LoRA | A collection for LoRA training and image generation on GPUs | You want to fine-tune FLUX.2 Klein and generate images |
| GROMACS molecular dynamics | Molecular-dynamics stages and scientific result visualization | You want to distribute protein simulation work |
| Gaussian Splatting pipeline | Video-frame extraction, structure from motion, GPU training, and point-cloud output | You need a multi-step 3D reconstruction workflow |
| GUI control showcase | Every OpenJD job-parameter GUI control and UI metadata option | You are designing a bundle submission interface |
| Houdini Husk USD render | USD dependency discovery and rendering with Husk/Karma | You need a concise USD render job with asset introspection |
| Infinigen scene generation | Procedural indoor and outdoor scene generation on GPU workers | You need synthetic photorealistic datasets |
| Job attachments input guide | Input path metadata and attached script files | You are learning how job attachment inputs are materialized |
| Job attachments output guide | Collecting declared job output files | You are learning how job attachment outputs are returned |
| Job development progression | A staged path from inline commands to a tested bundled Python package | You want to grow a maintainable job without starting complex |
| Daemon-process environment | Starting a background process once and sharing it across tasks | Application startup should be amortized within a session |
| Environment variables | Setting variables at job and step scope with OpenJD environments | Tasks need consistent runtime configuration |
| Environment-provided command | Creating a command and adding it to PATH for job steps |
Setup should expose reusable tooling to every step |
| KeyShot standalone | Frame-parallel KeyShot rendering on Windows | You render KeyShot scenes with the standalone interface |
| List available Conda packages | Querying a Conda channel from a Deadline Cloud job | You need to inspect packages visible to workers |
| Maya Arnold export and render | Exporting .ass once, then rendering frames with Arnold kick |
You want separate DCC export and renderer-only steps |
| Maya CLI render | Rendering a Maya scene with the CLI Render command |
You need a small Maya command-line example |
| Monte Carlo simulation | Parallel financial simulation followed by result aggregation | You want a non-rendering fan-out/fan-in workload |
| MuJoCo sim-to-policy | Simulation data generation, policy training, and rendered evaluation | You need a multi-step robotics ML workflow |
| Nuke render | Frame-parallel headless compositing with nuke -x |
You need to render Nuke scripts on workers |
| Pip package job | Declaring Python dependencies for a pip queue environment | A shared queue environment should provide job packages |
| Pip self-contained job | Creating and activating a pip environment inside one bundle | You cannot or do not want to configure the queue |
| POV-Ray 3.7 | Raytracing with a Conda-provided command-line renderer | You want a portable, lightweight render example |
| Redshift 2025 | Rendering Cinema 4D Redshift scenes with redshiftCmdLine |
You need direct Windows Redshift command-line rendering |
| Satellite classification | Per-tile image classification followed by mosaic assembly | Independent input files should fan out and merge |
| Simple job | The smallest developer-guide OpenJD job bundle | You are submitting your first custom job |
| SSH to SMF | Temporary Linux SSH access through an SSM hybrid managed node | You need interactive debugging on a service-managed worker |
| SSH to SMF on Windows | Temporary RDP, SSH, or PowerShell access through SSM | You need interactive debugging on a Windows worker |
| Task chunking | A collection of contiguous and non-contiguous chunking patterns | Per-task startup overhead should be shared across frames |
| Maya tile render blog sample | The adaptor customization and tiled-render workflow from the AWS blog | You are following the tile-rendering walkthrough |
| Maya Arnold tiled render | A three-dimensional task space and FFmpeg tile assembly | Arnold images should render as distributed tiles |
| Maya V-Ray tiled render | V-Ray tile rendering followed by OpenImageIO assembly | You need tiled EXR output from Maya and V-Ray |
| V-Ray Linux region render | Region rendering, asset discovery, path mapping, and image merge | You render .vrscene files in parallel regions on Linux |
| Maya Arnold turntable | Building a scene around an OBJ, then rendering frames and encoding video | You need an easy-to-submit 3D asset review utility |
| Studio VFX shot pipeline | A launcher, static render bundle, Conda software delivery, and Flow publishing mapped onto a studio's existing pipeline | You are porting a studio shot pipeline to Deadline Cloud |
| AutoDock Vina virtual screening | Parallel molecular docking and ranked result aggregation | You want to screen many ligands against a protein target |
| vLLM evaluation leaderboard | Parallel model evaluation and final CSV/Markdown aggregation | You need to compare multiple LLMs across benchmarks |
| V-Ray standalone render | Rendering with a Conda-provided V-Ray executable | You need a basic standalone V-Ray bundle |
| VRED render | Headless VRED rendering, tiling, and Python API control | You render VRED scenes with VRED Core or Pro |
| VTK visualization | Running a VTK Python visualization script | You need a portable scientific visualization job |
These compact bundles are intended to be read alongside reference documentation. The simple job accompanies the developer guide's developer farm walkthrough. The environment variables, environment-provided command, and daemon-process environment bundles accompany Control the job environment and demonstrate OpenJD environments; see the queue environment samples for session-wide alternatives.
The job attachments input and output bundles accompany the developer guide's
job attachments walkthrough. They show how OpenJD
PATH parameter data-flow metadata and asset_references.yaml jointly describe inputs and outputs, allowing the same bundle patterns to work with
job attachments or shared filesystems. The GUI control showcase is the compact reference for every control
supported by OpenJD job-parameter UI metadata.
With a job bundle in hand, the Deadline Cloud CLI provides ways for you to submit jobs to run on your Deadline Cloud queues. Read more about how to submit a job in the Deadline Cloud developer guide.
Here's the submitter GUI you can see after configuring the Deadline Cloud CLI
and running deadline bundle gui-submit blender_render/ in this samples directory:
Alternatively, you can submit this job bundle with the command
deadline bundle submit --name Demo -p BlenderSceneFile=<location-of-your-scene-file> -p OutputDir=<file-path-for-job-outputs> blender_render/
or use the deadline.client.api.create_job_from_job_bundle function in the deadline Python package.
If you do not want to use the deadline Python package's support for features like job attachments, you can also submit the job template by calling the
deadline:CreateJob API directly.
The Houdini Husk USD render sample shows how to use the Houdini Husk CLI USD renderer using a short job template and service-provided Conda packages. It also shows how to write a custom asset introspection tool for job attachments, ensuring that only the required data is uploaded while removing manual steps for artists.