Kinora: A Versioned Workflow for Text-to-Motion Generation, Robot Retargeting, and Interactive Validation
Kinora: A Versioned Workflow for Text-to-Motion Generation, Robot Retargeting, and Interactive Validation
This research presents Kinora, a system integrating text-to-motion generation, motion libraries, local agent tools, robot retargeting, and trajectory refinement via versioned artifacts. Evaluated on server 0.5.1 and Gadget 0.15.2, the system passed numerical gates in four CUDA cases while retaining warnings for foot drift in turning splices; simulation replay succeeded on a positive fixture but was quarantined for a suspended control. The authors note these results support integration contracts without establishing semantic improvements, broad feasibility, or safety.
The paper introduces Kinora, a versioned workflow system for text-to-motion generation and robot retargeting, evaluated on specific server and Gadget versions with regression tests that confirm numerical gate passage but explicitly do not validate semantic quality, population-level feasibility, or hardware safety.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
regression test executions
52
View original evidence
The current release uses an independently specified 52-joint authoring rig and procedural skeleton previews, separating the distributed human geometry from externally installed generation models and robot assets. Human-motion edits preserve frames outside a selected interval, while robot edits create immutable candidates whose reports govern acceptance and export. We evaluate server 0.5.1 and Gadget 0.15.2 with 156 c
Open source S3support-foot drift proxy warning
14.3
View original evidence
ore regression test executions, 49 Gadget executions, four CUDA generation/editing cases, a 30-frame G1 retargeting check, and a 3,600-frame synthetic G1 fixture paired with a suspended negative control. All four CUDA cases passed their numerical gates; the turning splice nevertheless retained a 14.3 cm support-foot drift proxy warning. The G1 positive fixture completed bounded simulation replay, whereas the suspende
Open source S4
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
The current release uses an independently specified 52-joint authoring rig and procedural skeleton previews, separating the distributed human geometry from externally installed generation models and robot assets. Human-motion edits preserve frames outside a selected interval, while robot edits create immutable candidates whose reports govern acceptance and export. We evaluate server 0.5.1 and Gadget 0.15.2 with 156 c
Open source S3
ore regression test executions, 49 Gadget executions, four CUDA generation/editing cases, a 30-frame G1 retargeting check, and a 3,600-frame synthetic G1 fixture paired with a suspended negative control. All four CUDA cases passed their numerical gates; the turning splice nevertheless retained a 14.3 cm support-foot drift proxy warning. The G1 positive fixture completed bounded simulation replay, whereas the suspende
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org