MULTIRENDER
Blender logo
For Blender Users

HoloMARI MultiRender for Blender

WHAT IS MULTIRENDER?

MultiRender is a Blender add-on that launches multiple headless Blender workers—typically one (or more) per GPU—to render frames or HoloMARI camera jobs in parallel. It brings stable GPU pinning (UUID/Bus), a lightweight job scheduler, per-worker logs, and clean cancellation. Goal: finish renders sooner by saturating every available GPU.

WHY USE THIS ADD-ON?

• Finish sooner: split frames or MARI camera jobs across parallel workers.
• Rock-solid GPU binding: avoid “ghost” devices via UUID/Bus mapping.
• Resilient jobs: automatic retries & requeue on worker errors.
• Operational visibility: per-worker logs; one-click “Open Logs.”
• HoloMARI-native: dispatch FRAME/CIRCLE jobs and export .mari3d on completion.
Built for multi-GPU workstations and creators who value time.
DOWNLOAD MULTIRENDER
Get the Blender add-on to render across multiple GPUs and instances in parallel.
DOWNLOAD MULTIRENDER ADD-ON

WHO IS IT FOR?

• Blender artists pushing animation / image sequences.
• HoloMARI creators exporting FRAME or CIRCLE camera jobs.
• Studios with multi-GPU rigs needing predictable device pinning and high throughput.

RENDER MODES

• FRAMES: distribute frame numbers across workers (use image sequences; video formats are blocked).
• MARI: dispatch per-camera H/V index jobs (FRAME or CIRCLE); export a unified .mari3d on finish.

HOW IT WORKS

1) Manager (in your Blender session): choose GPUs & instances per GPU, write a temp job, and launch workers.
2) Workers (headless blender -b): each pins to one GPU, asks the scheduler for work, renders, and reports status.
3) Completion: logs saved; for MARI jobs, a final .mari3d export is written to your source blend folder.

FEATURE HIGHLIGHTS

• Multi-GPU, multi-instance dispatch (DYNAMIC / STRIDE).
• Deterministic device selection (UUID/Bus) + optional ghost-device filtering.
• Per-worker log folders & an Open Logs action.
• Clean cancel with graceful teardown (ESC).
• Consistent output paths rebased to the original .blend directory.
• Works headless — perfect for render nodes.

COMPATIBILITY & REQUIREMENTS

• Blender: 3.x+
• GPUs: NVIDIA recommended for CUDA / UUID pinning.
• OS: Windows / Linux (headless supported). macOS GPU parallelism may vary.
• HoloMARI: required only for MARI mode; not needed for plain FRAMES mode.
• Output: Image sequences for FRAMES mode.

QUICK START

1. Install: Edit ▸ Preferences ▸ Add-ons ▸ Install… → select “HoloMARI_MultiRender.py” → enable.
2. FRAMES: set Render Output to an image format (PNG, EXR, etc.).
3. MARI: configure HoloMARI cameras (FRAME or CIRCLE).
4. Open Render Properties ▸ Multi-GPU Frames / Multi-Instance MARI and pick GPUs + instances.
5. Click Render (Multi-GPU Frames) or Render (Multi-Instance MARI).
6. Use Open Logs. Press ESC to cancel cleanly if needed.
FAQ

MultiRender FAQ

Q: Does this replace Blender’s built-in queue?
A: It complements it. MultiRender runs parallel headless instances with stable device pinning and HoloMARI job dispatch for maximum throughput.
Q: Can I run multiple instances on a single GPU?
A: Yes. Use “instances per GPU.” It can help latency-hidden workloads. Test to find the sweet spot for your scene.
Q: Why block video formats in FRAMES mode?
A: Parallel workers can contend for encoders and file handles. Image sequences are safer and more robust across processes.
Q: What if a worker crashes?
A: The scheduler retries and requeues the failed unit. Review its specific log in the logs folder.