Supported data types

We support the following data types natively:

Content Type	File Extension	Max File Size
Images	.jpg .jpeg .png .tiff .tif	1 GB
Images (medical formats)	.dcm	100 MB
Videos	.mp4	1 GB
Text	.csv .txt	100 MB
Audio	.wav, .mp3, .ogg	100 MB
PDF	.pdf	100 MB

❗️
S3 Imports have a maximum size limit of 5GB

We currently support the following DICOM syntaxes:

Implicit VR Little Endian
Explicit VR Little Endian
Explicit VR Big Endian
JPEG Baseline (Process 1)
JPEG Lossless, Nonhierarchical (Processes 14)
JPEG Lossless, Non-Hierarchical, First-Order Prediction (Process 14 election Value 1])
)
JPEG 2000 Image Compression (Lossless Only)
JPEG 2000 Image Compression
RLE Lossless

If you use a different syntax, please inform your project manager so we can prepare our systems accordingly!

The following data types require intervention from our data engineers:

Content Type	File Extension
Images (medical formats)	.nii, .svs
Videos	.avi, .mov
Text Only for Centaur-hosted, it needs to be converted to CSV for upload to the portal or submitted through API Self-hosted, no processing is needed	.txt

Working with long assets. Some assets are too long to label in a single pass. For video, Centaur can split a file into shorter segments for labeling, set up per task. For other content types, split the files into shorter units before you import them, and size each unit so that a labeler has enough context to answer the prompt on its own. Talk to your project manager if you are not sure how to divide your data.

Audio: Legacy Workflow

If you have existing workflows that use video-based audio annotation, you can convert .wav or .mp3 files to .mp4 using the audio-to-video conversion recipe. However, native audio annotation is now the recommended approach for new audio tasks.