In the digital age, where consumer electronics, healthcare, surveillance, and cloud computing have significantly increased the amount of image-based data, image security has become an issue of great concern. Traditional encryption techniques are unable to provide multiple levels of security, computational efficiency, and verifiable lossless recovery. This study explains the Hybridcrypt-AE, a new image encryption method built upon a hybrid autoencoder (hybridae) to enhance deep-feature key augmentation using plainimage dependent dynamic key generation via SHA-512, dual stream PCG64 based on 6D hyperchaotic PRNG, fisherman-yates pixel permutation and three rounds of dual feedback permutation-diffusion cipher. The scheme achieves a NPCR of 99.61% , 33.45% UACI, 7.9998 bits of information entropy, 9.07 dB of PSNR, and decrypted PSNR greater than 100 % with an SSIM of 1.0. The key space is at least 63,000 times larger than all of the other methods reviewed in this paper with respect to exponent, at 2^2097664 . Encrypting a 512 x 512 file uses 0.2513 seconds while Deepkey-tpe uses 0.728 seconds to encrypt. All of the individual components have been shown to add quantitatively towards adding security with ablation studies. The comparison of the state-of-the-art shows that Hybridcrypt-ae has better security, efficiency and recovery quality.