Worqgrid Studio
The annotation studio, built into the platform.
Text, image, audio, and video, all first-class, plus RLHF preference ranking built into the text annotator. No bolt-on integrations, no separate logins. Your annotators open Worqgrid and get to work.
Text annotator
Classification, span/NER, RLHF preference ranking, and free-form rewrites. Custom label schemas with hotkeys, hierarchical labels, and per-span notes.
Image annotator
Bounding boxes, polygons, keypoints, and per-pixel segmentation. Multi-class with hierarchical taxonomies and instance tracking.
Audio annotator
Transcription, speaker diarization, and event tagging on a multi-track waveform. Region labels with millisecond precision.
Video annotator
Frame-level boxes with object tracking, action segments, and timeline-scrubbed labels. Keyframe interpolation included.
Quality first
QA built into every step
Every Studio annotator, text, image, audio, video, ships with the same QA primitives you'd otherwise build yourself.
Live IAA tracking
Cohen's κ and Krippendorff's α computed per annotator, per task, in real time. Trigger calibration sessions when scores drift.
Blind double-pass review
Route every Nth item to two annotators independently. Disagreements escalate to a reviewer queue automatically.
Gold-set sampling
Inject pre-annotated gold items at a configurable rate. Auto-flag annotators whose gold accuracy slips below threshold.
Versioned guidelines
Edit rubrics with full history. Re-run trial tasks against new guidelines without breaking the live queue.
Reviewer escalation
Three-role workflow (annotator → reviewer → PM). Items can be returned, overridden, or sent back for re-annotation.
Throughput dashboards
Per-annotator items/hour, accept rate, and median review time. Export to CSV or pull via the REST API.
Four modalities, one workspace.
No more shipping a CVAT for vision and a doccano for text and a separate spreadsheet for RLHF rankings. Your team works in one place.