<p>Bitcoin volatility has posed significant challenges to investors, making it a focal point for researchers. With increasing use of the internet, daily news is considered to have a substantial influence on bitcoin prices. Traditional methods relying solely on technical indicators or social sentiment often fail to capture the nuanced relationships between these factors, leaving a significant gap in accurately predicting price movements. In this research, we predict bitcoin prices by leveraging news data extracted from Common Crawl News dataset, in conjunction with bitcoin prices obtained from Coinbase Application Programming Interface. We propose a multimodal deep learning model named, Generative Pre-Trained Transformer 3.5 (GPT3.5) enhanced Convolutional Neural Network (CNN) Positional encoding-based Transformer Encoder (GPT-CNN-PTEN), designed to predict Bitcoin prices during both bullish and bearish market phases. To achieve this, we employ transformer encoder layers with CNN positional encoding to capture relations in the text and bitcoin prices and handle temporal dependence associated with bitcoin prices. The word embeddings generated using the Ada002 model from the text summarized by GPT3.5 turbo model captures the context of lengthy news articles into concise and meaningful inputs to the model. The dataset from October 2022 to January 2023 was used to train the model. The model was tested on three distinct datasets representing various market conditions. The results are noteworthy achieving mean squared error of 0.001, mean absolute percentage error of 7%. Moreover, the model demonstrated its predictive power by accurately forecasting a bull run hours before it occurred in January 2023.</p>