Skip to content
NAMEFRAMECommercial PreviewApply for Pilot
EasySynth

NameFrame and EasySynth

These two are often mentioned in the same sentence and they do different jobs. EasySynth renders image sets from a camera sequence you keyframe. NameFrame builds a labelled dataset and then checks it. If what you need is depth and normals from a path you have already animated, the first one is a shorter road.

EasySynth engine
5.5 to 5.7
NameFrame engine
5.8
EasySynth licence
MIT
Verified
27 August 2026

They were built for different jobs

EasySynth is a rendering plugin. You place cameras, put them in a Level Sequence, pick which targets you want, and it writes colour, depth, normal, optical flow and semantic segmentation images along that path. It is direct, it is open source, and for a lot of work that is the entire requirement.

NameFrame starts one step later. It assumes the images are the easy part and that the expensive question is whether the labels are right: which object is which, how much of it you can see, and whether anything in the run contradicts itself. So it renders an identity pass alongside the picture, measures boxes and masks from it, and grades the result before it will export.

Neither of those is an argument against the other. If your dataset is camera-sequence driven and you want optical flow, EasySynth does something we do not do at all.

Side by side

NameFrameEasySynth
What it is forGenerating labelled computer-vision datasets end to endRendering image sets from a camera sequence
Unreal versions5.8, verified on 5.85.5 to 5.7
Colour imagesYesYes
DepthMetric, float32 metres per pixelYes
Semantic segmentationYesYes
Instance segmentationYesNot documented
Bounding boxesMeasured from the identity passNot documented
Occlusion-aware labelsVisible pixel share per instance, and a gate on itNot documented
YOLO and COCO exportYOLO and COCO, plus the raw dumpCamera poses as CSV, rig as ROS-format JSON
Built-in randomizationSpawn zones, exclusion zones, materials, animations, entity swappingNot documented
Camera samplingSampled under constraints, with validation and retriesA Level Sequence you keyframe
Deterministic runsSeeded, and a run can be replayedNot documented
Dataset validationGraded and gated before export, with a published reportNot documented
CLI and automationCLI, REST and MCPNot documented
In-editor interfaceStudio panel in the editorEditor plugin
LiDAR and point cloudsExperimental raycast LiDARNot documented
LicenceCommercial previewMIT
MaturityNew. Not on sale yetEstablished, open source

Last verified 27 August 2026, from EasySynth on GitHub. Where their documentation does not answer a question, this page says so rather than counting it as a missing feature. If something here is out of date or wrong, tell us and it gets corrected.

Which one to reach for

Choose NameFrame if

  • you need bounding boxes or instance masks and you do not want to write the code that derives them
  • occlusion matters, because a person behind a fence has to be labelled as partly visible rather than as fully present
  • you want the same seed to give you the same dataset, so an ablation holds everything else still
  • you want YOLO or COCO out of the run rather than a conversion script
  • you are on Unreal 5.8

Choose EasySynth if

  • you are on 5.5, 5.6 or 5.7 and cannot move
  • you want optical flow or surface normals, which NameFrame does not produce
  • your captures are driven by a Level Sequence you have already built
  • you want something MIT-licensed you can read and change
  • you are happy to build the annotation and validation layer yourself

What this page is not

This is written by the people who make one of the tools on it, so read it that way. What we can do is keep it checkable: every row came from the other project’s own repository or documentation on the date at the top, the link is there, and a question their docs do not answer is written as not documented rather than as a missing feature.

We also have one end-to-end result and it is not flattering. A detector trained only on NameFrame frames, with no real images anywhere in training, improved on its synthetic test split and lost recall on real drone footage. It is published with the failure pictures on the experiment page. No comparison table on this site should be read as evidence that our output trains a better model, because we have not shown that.