We present an experiment conducted on the augmentation of older grayscale datasets designed for automatic text recognition on contemporary handwriting (IAM-Database). The augmentation method relies on the addition of colored backgrounds taken from real-world historical blank pages and allows us to create an enhanced version of IAM-Database. We train various transcription models playing on the composition of trainset and validationset using the original and enhanced IAM-Database. We test the resulting models against the original and enhanced testsets, as well as a testset composed from real-world historical documents. We find that though the transcription engine proves robust to color changes, this technique could be used to bring up to speed older grayscale datasets to create transcription models efficient on historical handwriting. Additionally, we consider the environmental costs of using enhanced data as opposed to the original dataset, and find that the impact is minor.