<p>This paper introduces QuanFlex-GAN, a flexible hybrid quantum-classical generative adversarial network for image generation using quantum generative modeling. The proposed generator combines two parameterized quantum circuits, one processing input noise and the other processing label information, with a lightweight classical neural network to synthesize high-quality images. QuanFlex-GAN is capable of generating diverse handwritten digits, including single digits, digit sequences, and randomly sampled digits, demonstrating strong generative flexibility and controllability. The discriminator adopts a standard architecture to ensure stable and efficient training. Experimental results show that QuanFlex-GAN outperforms both classical CGANs and existing QGANs in terms of image quality and training robustness. This work highlights the potential of quantum generative models in developing more adaptable and expressive image generation frameworks.</p>