# Diff between trocr-printed and trocr-handwritten
**URL:**
**Author:**  [@Kforcode](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FKforcode)%5C
**Post date:** [November 24, 2021, 11:41am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fdiff-between-trocr-printed-and-trocr-handwritten%2F12239%2F1 "2021-11-24T11:41:26Z")
what’s the Diff between trocr-printed and trocr-handwritten, other than the dataset they were trained on, because I inference an image with both the model and found handwritten gave correct output but printed missed a character
---
**Author:**  [@nielsr](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fnielsr)%5C
**Post date:** [November 24, 2021, 12:01pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fdiff-between-trocr-printed-and-trocr-handwritten%2F12239%2F2 "2021-11-24T12:01:12Z")
Hi,
There’s no difference other than the datasets they were fine-tuned on. TrOCR-printed will be better at printed text (see the images in the [SROIE dataset](https://arxiv.org/pdf/2103.10213.pdf)), while TrOCR-handwritten will be better at handwritten text (see some [sample images](https://fki.tic.heia-fr.ch/databases/iam-handwriting-database) of the IAM dataset).
---
**Author:**  [@Kforcode](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2FKforcode)%5C
**Post date:** [November 24, 2021, 12:02pm UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fdiff-between-trocr-printed-and-trocr-handwritten%2F12239%2F3 "2021-11-24T12:02:58Z")
so the handwritten model prediction being correct and printed model being incorrect would be random