State-of-the-art speech synthesis models try to get as close aspossible to the human voice. Hence, modelling emotions is anessential part of Text-To-Speech (TTS) research. In our work,we selected FastSpeech2 as the starting point and proposed aseries of modifications for synthesizing emotional speech. According to automatic and human evaluation, our model, EmoSpeech, surpasses existing models regarding both MOS scoreand emotion recognition accuracy in generated speech. Weprovided a detailed ablation study for every extension to Fast-Speech2 architecture that forms EmoSpeech. The uneven distribution of emotions in the text is crucial for better, synthesized speech and intonation perception. Our model includes aconditioning mechanism that effectively handles this issue byallowing emotions to contribute to each phone with varying intensity levels. The human assessment indicates that proposedmodifications generate audio with higher MOS and emotionalexpressiveness.