Houdini Karma AOVs: Render Vars, Multi-Channel EXR and Cryptomatte
Karma writes AOVs (render vars) into one multi-channel EXR per frame by default. What to get right before a long Karma XPU render:
- Create: Karma Render Settings for standard AOVs; Extra Render Vars and Karma AOV VOP for custom ones
- Precision: 16 bit for colour passes, 32 bit for position and depth
- Crypto: Karma Cryptomatte LOP, IDs from prims, materials or a string primvar
- Cost: each AOV adds frame-buffer memory; many AOVs slow Karma XPU
- H22: one file per render var, separate _denoise AOVs, a global AOV material
A beauty pass alone rarely survives compositing. Lighting, utility and ID passes let a compositor rebalance lights, fix a material or pull a matte without sending the shot back to rendering. In Solaris these passes are USD render vars, collected into render products, and Karma writes them to EXR.
This guide explains the ways to create AOVs in Houdini, how they reach the EXR, how to set up Cryptomatte, and what many AOVs cost in memory and speed on Karma XPU, with the changes Houdini 22 brings.
Checked against the SideFX Houdini 22 documentation (AOVs, Cryptomatte, Karma XPU, Inside husk, What’s new in Karma), November 2026.
Four ways to create AOVs in Solaris
SideFX calls the Karma Render Settings LOP the most convenient way: its Image Output ▸ AOVs (Render Vars) tab lists every built-in AOV. The other nodes exist for when you need more control.
Four ways to create AOVs in Solaris
From the Houdini 22 AOV workflow page.
| Method | Where | Use it for |
|---|---|---|
| Karma Render Settings LOP | Image Output ▸ AOVs (Render Vars) tab | Most shots: tick the standard AOVs you need |
| Karma Standard Render Vars LOP | Separate node | Per-AOV precision (16 or 32 bit), filters and output colorspace |
| Extra Render Vars / Additional Render Vars LOP | Render settings section or separate node | Custom AOVs, for example your own light path expressions |
| Karma AOV VOP | Inside a material, Create Render Var on | Material outputs such as a mask or a texture layer |
| Karma Cryptomatte LOP | Separate node | ID mattes from prims, materials or string primvars |
The Karma Standard Render Vars LOP offers the same categories as the render settings node but adds per-AOV parameters, including Precision with 16 Bit or 32 Bit. A common choice is 16 bit for colour passes, which halves their size on disk, and 32 bit for data passes such as position and depth, where small errors show up in compositing.
Sources: AOVs, Karma Standard Render Vars.
How render vars reach the EXR
When you render through the USD Render ROP, the active AOVs are saved in a multi-channel EXR file. A Render Product Edit LOP placed after the render settings controls the channel order through Ordered Render Vars, and the Karma Render Products LOP creates several render products that share common settings, for example the same AOVs from two cameras.
Houdini 22 adds Split Render Products by Render Var, which writes a separate file per render var. That helps when a compositing package or a review tool reads single-layer files more easily, at the cost of many more files per frame. Other Houdini 22 additions: denoised AOVs can be written as separate AOVs with a _denoise suffix, and a Global AOV material replaces editing every shader to add the same AOV.
Sources: AOVs, What’s new in Houdini 22: Karma.
Cryptomatte in Karma
Cryptomatte gives the compositor an ID matte for every object or material without rendering separate mask passes. Karma XPU has supported Cryptomatte since Houdini 20, and Karma builds the IDs from three sources: primitives, materials bound to prims, and constant or uniform string or integer primvars.
Use the Karma Cryptomatte LOP to set it up, or add a render var by hand in Extra Render Vars and turn on its Cryptomatte checkbox. For instanced scatters, SideFX’s advice is to assign a unique value of a string primvar to each instance and point the Cryptomatte render var at that primvar; no material stylesheets are needed.
Before the full render, open one frame in your compositing package and check that the Cryptomatte layers and their manifest are present and that picking an object returns the expected name.
Sources: Cryptomatte and material AOVs in Karma, What’s new in Houdini 20: Karma.
For scatters, a string primvar per instance is the most flexible source: each copy gets its own matte without adding materials or prims.
Light-group AOVs with Split per LPE Tag
To rebalance lights in compositing, give each light an LPE tag on the Karma tab of the light (or with the LPE Tag LOP) and turn on Split per LPE Tag on the AOVs you want split. With tags Key, Fill and Rim, the beauty AOV C becomes C_Key, C_Fill and C_Rim.
Every tag multiplies every split AOV, so the count grows fast. Tag only the lights you plan to grade separately, and split only the AOVs the compositor will use. Our guide to light groups and light linking in Karma covers custom light path expressions in more detail.
Sources: Light Path Expressions, AOVs.
What many AOVs cost on Karma XPU
AOVs are not free. Karma allocates render vars in VRAM and RAM, and SideFX notes that a scene with 15 AOVs needs at least 1.5 GB for frame buffers alone; larger resolutions need more. Because Karma XPU does not pool memory across GPUs, that buffer memory is needed on every card, on top of the scene.
The Karma XPU page also warns that many AOVs can slow XPU down, and that only the listed sample blending modes are supported on AOVs. When a shot needs many AOVs, SideFX suggests raising KARMA_XPU_NUM_PER_DEVICE_BLENDING_THREADS from its default of 1 to between 2 and 4:
Two more limits worth knowing: deep output from Karma XPU requires tiled rendering because XPU has no bucket mode, and the denoiser on the Filters tab only processes the render vars listed in its AOVs parameter. For memory planning on 32 GB cards, see how much VRAM Houdini GPU rendering needs.
Sources: Karma XPU, Inside husk.
Because the buffers are allocated on every GPU, the safe way to plan is per card: scene plus all AOV buffers at the delivery resolution must fit inside one GPU’s memory.
SaaS or IaaS render farm for AOV-heavy Karma renders?
Which farm type fits an AOV-heavy shot
Both render AOVs; they differ in how easily you can check the EXR before the full run.
| Situation | Usually a better fit | Why |
|---|---|---|
| Standard AOV set, layout already approved by comp | SaaS (managed farm) | Submit and collect the EXRs |
| You need to check Cryptomatte and channel names on the render machine first | IaaS (remote workstation) | Render one frame, open it, fix, render again in the same session |
| Many AOVs close to the VRAM limit | IaaS with 32 GB cards | You can watch memory and adjust blending threads or AOV count |
| Deep output or tiled stills | IaaS | You control tiling, stitching and denoising |
| Small job, few AOVs | SaaS | No setup time |
Rule of thumb: if the AOV list is fixed and tested, a managed farm is the least work. If the shot sits close to the VRAM limit or you are still adjusting AOVs and Cryptomatte, a machine you control saves round trips.
Rendering AOVs on an iRender server
On iRender you rent a whole Windows workstation with your own Houdini and compositing software, so you can render a frame and check it in the same session:
- Upload the project, with USD, textures and caches, through iRender Drive.
- Start a 4x RTX 5090 server (32 GB per card), which we recommend for Karma XPU; 8x servers also render Karma XPU, but it does not get proportionally faster.
- Render one frame with the full AOV list and open the EXR to check channels, precision and Cryptomatte.
- If AOVs slow the frame down, try a higher KARMA_XPU_NUM_PER_DEVICE_BLENDING_THREADS or drop unused AOVs.
- Render the range and download the EXRs from iRender Drive.
AOV checklist before the full render
Use this as a final check before you start the full range on any machine.
Frequently asked questions
1. How do I add AOVs to a Karma render in Houdini?
Tick them on the Karma Render Settings LOP under Image Output ▸ AOVs (Render Vars). Use Extra Render Vars, the Additional Render Vars LOP or the Karma AOV VOP for custom AOVs.
2. Does Karma XPU support Cryptomatte?
Yes. Cryptomatte has been available for Karma XPU since Houdini 20. Set it up with the Karma Cryptomatte LOP.
3. Can Karma write each AOV to its own file?
Yes, from Houdini 22: Split Render Products by Render Var writes a separate file per render var. By default all active AOVs go into one multi-channel EXR.
4. Why is my Karma XPU render slower with many AOVs?
SideFX notes that many AOVs can slow XPU down. Remove unused AOVs or raise KARMA_XPU_NUM_PER_DEVICE_BLENDING_THREADS to 2–4.
5. How much memory do AOVs use?
SideFX gives an example: a scene with 15 AOVs needs at least 1.5 GB for frame buffers alone, more at higher resolutions, and Karma XPU needs it on every GPU.
6. Should AOVs be 16 or 32 bit?
The Standard Render Vars LOP offers both. 16 bit is usually enough for colour passes; use 32 bit for position, depth and other data passes.
7. How do I control the order of AOVs in the EXR?
Add a Render Product Edit LOP after the render settings and set Ordered Render Vars.
8. Where can I see which AOVs a render will write?
All active AOVs are listed in the Scene Graph Tree under /Render/Products/Vars.
Render your Houdini Karma XPU scenes on iRender
iRender is an IaaS GPU render farm: you rent a bare-metal server with 4x NVIDIA RTX 5090 (32 GB) GPUs, an AMD Ryzen™ Threadripper™ PRO CPU, 256 GB ECC RAM and a 2 TB NVMe SSD, connect by remote desktop and work on it like your own workstation, with full administrator access. You install Houdini and use your own SideFX license; your server image is saved between sessions, so Houdini, your packages and the Karma XPU kernel cache stay installed. For Karma XPU we recommend the 4-GPU servers.
Multi-GPU servers are billed per minute (1-GPU servers have a one-hour minimum), and sessions of 3 hours or more get 10% off. Our 4-GPU servers are liquid-cooled; in our own measurements the GPUs hold 60–70°C under sustained render load. Elite support is available 7 days a week by hotline, live chat, Skype, WhatsApp and email.
This video walks through rendering a Houdini Karma XPU scene on an iRender server in five steps:
New users get a 100% bonus on the first top-up (minimum $50). See the Karma XPU render farm page for server options, or download the iRender GPU app.
Sources: AOVs, Cryptomatte and material AOVs in Karma, Karma Standard Render Vars, Light Path Expressions, Karma XPU, Inside husk, What’s new in Houdini 22: Karma, What’s new in Houdini 20: Karma. Checked November 2026.
If Karma XPU renders are slow on your machine, iRender’s Karma XPU render farm gives you a dedicated remote server where you install Houdini, bring your own license and render on multiple RTX 5090 GPUs (for Karma XPU, 4 GPUs is usually the most efficient setup).


