Apart from the natural variation in the perception of language and interpretation, the AI image generators might be designed with specific parameters or algorithms that prioritize certain keywords or phrases, or they might not have been trained to recognize the distinction between “pedals” and “petals” in the context of describing flowering trees. The AI models are trained on large datasets of images and corresponding captions, so if the distinction between pedals and petals wasn’t present in the training data, it might not be recognized in the generated images. Also, free AI image generators might have certain limitations in their models or processing capabilities, which could affect their ability to generate images based on specific or nuanced prompts.