prompt_1 = "A Balrog in Moria high defination image" prompt_2 = "A still life DSLR photo of Sauron in Mount Doom" prompt_3 = "high quality painting of Gondor with fire raining from sky" prompt_4 = "Elves from Rivendell with Hobbits high quality image" interpolation_steps = 6 batch_size = 3 batches = (interpolation_steps**2) // batch_size encoding_1 = tf.squeeze(model.encode_text(prompt_1)) encoding_2 = tf.squeeze(model.encode_text(prompt_2)) encoding_3 = tf.squeeze(model.encode_text(prompt_3)) encoding_4 = tf.squeeze(model.encode_text(prompt_4)) interpolated_encodings = tf.linspace( tf.linspace(encoding_1, encoding_2, interpolation_steps), tf.linspace(encoding_3, encoding_4, interpolation_steps), interpolation_steps, ) interpolated_encodings = tf.reshape( interpolated_encodings, (interpolation_steps**2, 77, 768) ) batched_encodings = tf.split(interpolated_encodings, batches) outputs = [] for batch in range(batches): images.append( model.generate_image( batched_encodings[batch], batch_size=batch_size, ) ) images = np.concatenate(outputs) plot_grid(images, "lotr.jpg", interpolation_steps)