TTI-Eval is an open-source tool designed to evaluate text-to-image embedding models, such as CLIP, against custom datasets or those available on Hugging Face. It provides a straightforward and interactive evaluation process to estimate how well different embedding models capture semantic information within the dataset. TTI-Eval can be used by researchers and developers to select the most suitable model for their specific use case, improve the accuracy of natural language and image similarity search features, and curate datasets. The tool allows users to generate custom embeddings from model-dataset pairs, evaluate the performance of embedding models on custom datasets, and visualize the reduction of embeddings from two models on the same dataset. By using TTI-Eval, users can determine which model is ideal for their dataset and build active learning pipelines with Encord.