Getting Started With Olyphant Actor for Real Production Work
Olyphant Actor The Next Evolution In Digital Performance Technology is a digital performance pipeline tool that maps facial capture data onto 3D character rigs and generates voiced, animated performances from script input. It's been gaining traction in virtual production and indie game dev circles because it cuts the traditional motion-capture session down from a full-day shoot to something more like a few hours of setup plus render time. I've been running it on projects for about eight months now. First you export or record your actor's performance using a standard face-capture setup — Faceware, iPhone LiDAR, or whatever your budget allows. Olyphant Actor ingests that raw data, processes it through its retargeting engine, and then applies it to your target rig. The trick is that it doesn't just copy the animation. It re-maps the blendshapes to match the topology of your character, which means a human face gets translated into whatever mesh you're working with. Then there's the voice layer. The system can either use recorded audio you feed it or generate speech from a script using its built-in TTS module. You layer the audio under the animation and the system auto-syncs phonemes to the correct blendshape targets. I usually run a full session at about 120 frames per second on the capture side and downsample to 60 for the final output. Anything faster than that and you start seeing interpolation artifacts in the eye regions during quick head turns. That's a known issue and there's no official fix yet. My workaround was to keyframe the eye blink blendshapes separately and merge them back in after the main bake. Takes about twenty minutes per scene but it looks clean.
What Most People Get Wrong About This Pipeline
The biggest mistake I see people making is treating the retargeting pass as a one-click solution. The topology differences between a scanned human face and a stylized game character can cause serious overshoot in the jaw and cheek regions. I learned this the hard way on my first project — the character's mouth looked like it was stuck mid-chew through half the dialogue because the lip-sync weights were being applied to blendshapes that didn't correspond to the same anatomical positions. The fix is to manually remap at least thirty percent of the facial control nodes before you run the export. You spend another hour on that step but it saves three days of cleanup later. Another thing that catches people off guard is the TTS quality. The built-in speech generator sounds fine for background characters and NPC lines, but when you need emotional nuance — grief, sarcasm, whispered intimacy — it falls apart. The prosopagnosia effect kicks in where the voice is technically accurate but emotionally flat. I ended up recording my lead talent separately and using Olyphant Actor purely for the facial retargeting and secondary character voices. That split approach gives you the speed of automation where it matters and the quality of human performance where it counts.
Setting Up Your Rig for Compatibility
Before you even import anything, check that your character rig uses a standard ARKit or FLAME blendshape set. If it's built around a custom set like MetaHuman's default controls, you'll need to run the conversion script that comes with the package. It's located in the Tools directory and takes roughly ten minutes per character. Characters built on older systems like the FaceWare Human IK rig sometimes need manual bone mapping because Olyphant Actor's retargeting assumes a specific joint hierarchy. I had a pipeline artist spend an afternoon reworking a rig that turned out to be using legacy control names, and once we renamed the clusters to match the expected namespace, the retargeting worked immediately. Lighting matters more than you'd think if you're doing in-camera capture. The system can compensate for moderate lighting changes, but hard shadows across the face — especially under the eyes and nose — create data gaps that show up as jitter in the final render. I always make sure my capture talent has even frontal fill lighting. A simple two-point softbox setup at roughly equal intensity eliminates most of the jitter without requiring expensive gear.
Get the Full Details
Download and Installation Notes
You can find the latest build on the official Olyphant website. The download is approximately 2.4 gigabytes and requires at least 8 gigabytes of RAM dedicated to the processing plugin. Mac and Windows are both supported, though the rendering pipeline runs about fifteen percent faster on the Windows build according to my benchmarks. The installer places the plugin in your standard Unreal Engine or Unity plugin directory depending on which version you select during setup. There's a licensing step after installation where you enter your account key. The free trial gives you access to the full suite for fourteen days with a watermark on exports. After that it's a subscription model at around ninety-nine dollars per month or nine hundred and ninety dollars annually. For a small team working on a single project, the annual plan works out cheaper, but if you're only using this sporadically the monthly option is fine. No per-seat licensing friction which is unusual for tools in this space.
When This Tool Absolutely Will Not Work For You
If you need photorealistic human faces at feature-film resolution, this isn't your tool yet. The retargeting introduces subtle uncanny valley artifacts around the micro-expressions — things like the tiny lateral movements of the lips during consonant clicks or the asymmetric brow raises that happen during genuine surprise. These get caught on camera in close-ups and look wrong even if you can't immediately say why. For wide shots and medium framing it holds up reasonably well. I've used it for background performers in a series pilot and nobody flagged it during dailies. Same project, a close-up insert shot of the same performance, and the director asked if we had a real actor available instead. Also, the system struggles with extreme facial expressions. If your script calls for a full scream, a wail, or a grimace where the skin is deforming far beyond neutral, the retargeting tends to collapse the geometry. I got around this on one scene by blending a hand-keyed extreme expression with the automated retarget at the midpoint. The transition was seamless at twenty-fourths of a second of animation, which is exactly what that shot required. But that kind of intervention defeats the time-saving purpose of the tool in the first place. If you're doing high-end commercial work where every frame is scrutinized, you're better off sticking with traditional mocap and hiring a dedicated animator for cleanup. Olyphant Actor is strongest in the middle ground — projects that need quantity of performance without sacrificing baseline quality. It's a production accelerator, not a replacement for the craft.
