Vailabel Studio logoVailabel Studio
Offline-first · AI-assisted · Open source

Label your data with an offline AI copilot

Vailabel Studio is a local-first desktop studio for image, video, and multi-modal annotation. Detect, segment, and QA your labels with on-device AI — your data never leaves your machine.

  • Free & open source
  • Windows · macOS · Linux
  • No account required
Vailabel Studio · street-scene.jpg
personcartraffic light
AI copilot detected 4 labels
Features

Everything you need to label faster

A focused, native annotation studio — not a browser tab. Built for speed, privacy, and real ML workflows.

Every annotation tool

Bounding boxes, polygons, points, lines, circles, free-draw, and SAM smart-segment — each with a single-key shortcut.

Offline AI copilot

Ask it to detect, segment, suggest labels, or QA your work. Local ONNX models, human-in-the-loop, zero cloud calls.

Export anywhere

One click to LabelMe, COCO, YOLO (detection & segmentation), and Pascal VOC — ready for any training pipeline.

Local-first & private

Projects live in a local SQLite database on your machine. Nothing is uploaded unless you choose to.

Bring your own cloud

Optionally connect S3, Azure Blob, or Google Cloud Storage. Credentials are kept in your OS keychain.

Video & multi-modal

Frame-by-frame video labeling with object tracking, plus a multi-modal architecture spanning image, video, text & audio.

AI Copilot

An AI that labels with you

A chat copilot that sees the current image and takes labeling actions — running entirely on-device. It proposes; you approve. Nothing leaves your machine.

Detect

“Find all the cars” → bounding boxes from a local YOLO model.

Segment

Click an object → a clean polygon mask via MobileSAM.

Suggest labels

“What should I label here?” → one-click label chips.

QA review

“What did I miss?” → findings you approve or reject.

ONNX Runtime · optional CUDAWorks with LM Studio · Ollama · Jan
Read the copilot guide
AI CopilotOn-device
Detect all the cars in this image
Found 9 cars. Added them as predictions to review.
What did I miss?
2 unlabeled people on the left — accept to add them?
Ask the copilot…
Workflow

Fits the stack you already have

From first label to training-ready dataset — without forcing a new format or a cloud account on you.

Project types

Pick a task and the right tools light up.

Object detectionSemantic segmentationImage classificationKeypointsMixed shapes

Export formats

Train with the tools you already use.

COCOYOLOYOLO-SegPascal VOCLabelMe JSON

Storage

Local by default, cloud when you want it.

Local SQLiteAmazon S3Azure BlobGoogle Cloud Storage
Demo

See it in action

Watch the AI-assisted annotation workflow end to end.

AI-assisted annotation workflow
Roadmap

Shipped, building, and next

An honest view of where the studio is today and where it's headed.

Shipped

Full annotation toolset

Box, polygon, point, line, circle, free-draw

AI copilot

Detect, segment, suggest & QA labels

YOLO + MobileSAM

Local ONNX inference, optional CUDA

Multi-format export

COCO, YOLO, YOLO-Seg, VOC, LabelMe

Cloud storage

S3, Azure Blob & GCS buckets

Video annotation

Frame-by-frame with object tracking

In progress

Florence-2 engine

Captioning, OCR & open-vocab tasks

SAM 2

Higher-quality interactive segmentation

Dataset intelligence

Deeper quality & outlier insights

Conversational copilot

Multi-turn labeling dialogue

Planned

Grounding DINO

Open-vocabulary detection

Text & audio editors

NER, relations, ASR, segments

OCR as a task

First-class text recognition

Team collaboration

Multi-user projects

Start labeling in minutes

Download Vailabel Studio for free and put an offline AI copilot to work on your dataset today.