news

Beyond Vision: CVAT Goes Multimodal and Introduces Audio Annotation. Learn more

CVAT

Annotation & labeling for real‑world AI

The multimodal AI workspace for production‑grade, training‑ready data.

Jobs/Deck cam 3
snap_back_zonewinch · running
Frame 1 / 121

Ship higher-quality datasets, faster

Hand tools, AI models and annotators, in one seamless flow.

Images

One click masks an object. One more labels every match in the frame.

Drone survey frame of cattle grazing around a water trough

Interactors · SAM

0cattle

Video

Label a few frames. CVAT carries every label through the entire clip.

Frame 1 / 97

3D point clouds

Every cuboid sits beside its camera view, so you always know what you're labeling.

Perspective
camera_front.jpg

Audio

Speakers, speech and noise split on the waveform, with the transcript alongside.

bed12_night.wavS1 ClinicianS2 Patient
0:000:060:120:180:24

Transcript

  • 00:00Clinician: Take a deep breath in for me.
  • 00:17Patient: Sorry. It's always worse at night.

Your workflow, orchestrated

Your team, your models and external annotators, coordinated through the stages you define.

Projects/Downtown drive/Jobs

Annotation3

Validation1

Acceptance1

Job #16External
Job #15MCMaya C.
Job #14YOLO
Job #13DLDan L.
Job #12ARAna R.
  • Define the workflow

    Stages, labels and rules, set once per project.

  • Bring in anyone

    Your team, your models or external annotators, on the same jobs.

  • Stay in control

    Assign, track and accept every job from one place.

CVAT is how ML teams ship real-world AI.

Autonomous yard truck in a dark depot, lit by low sun under a half-raised door
CVAT