We haven’t tested Chandra yet, because it’s very new. Under the hood Tensorlake is very similar to Marker - it’s a pipeline based OCR API, we do layout detection, Text Recognition and Detection, Table Structure Understanding, etc. We then use VLMs to enrich the results. Our models are much bigger than marker, and thus takes a little longer to parse documents. We optimized for accuracy. We will have a faster API soon.
Comments
We haven’t tested Chandra yet, because it’s very new. Under the hood Tensorlake is very similar to Marker - it’s a pipeline based OCR API, we do layout detection, Text Recognition and Detection, Table Structure Understanding, etc. We then use VLMs to enrich the results. Our models are much bigger than marker, and thus takes a little longer to parse documents. We optimized for accuracy. We will have a faster API soon.