In the era of digitization, quick usage of information has transformed the environment of journalism, requiring novel techniques for content growth. The increase in social media platforms has enlarged the need for appropriate and intriguing news segments that interact with the audience. Through images shared on news sources and networking sites, there is a massive gathering of visual substance through the capacity to inform and improve textual tales. Enabling a sense of the visual facts turning into sensational content is a difficult feature. To solve these issues, this study proposed an Automatic News Article Creation (ANAC) from images to text. The proposed task makes use of contemporary developments in deep learning (DL) image advanced text creation tools, as well as captioning to offer a novel multimodal algorithm that automates the formation of articles from images. Collect a diverse set of images from different platforms, such as news websites, image repositories, and social media. Gaussian filtering (GF) was used as a preprocessing tool for the acquired data. Enhanced recurrent neural network (ERNN) and residual network-50 (ResNet-50) algorithms instigate through extracting the position from negotiable image file set-up data in the image. After image data extraction, the corresponding news headline is created. The headlines are utilized to generate a comprehensive news article with bidirectional encoders indicating from transformers (BERT). The proposed method outperforms diverse estimations and reaches 88 parameters. The findings of this investigation illustrate the ANAC model's higher performance. After 100 training epochs, the model had an accuracy of 0.88 with a categorical cross-entropy loss of 0.25, far exceeding traditional models. According to a comparison investigation, the proposed model achieved a BLEU-1 score of 0.88 and a ROUGE score of 88, outperforming T5 (47), PEGASUS (42), BART Modified (40), & Pipeline-BART (38) the model's ability to produce coherent and reliable news articles.