Yeah This is one I reason why I strongly believe in synthesing the music at runtime rather than at design time; with pre-rendered music you often get a combination of low quality, short music, few tracks. On other hand, hours long high fidelty 6.1 audio fits perfectly in a few mb if synthesized at runtime.

If the wav files are high quality 44100 16-bit files, convert them to ogg. Otherwise, convert them to 22150 Hz 4-bit ADPCM audio, for sound effects this is acceptable. For the ogg music, don't go below 128kbps, and use VBR encoding.

png files can be converted to 8-bit colour, while keeping transparency. Especially if they are small, this often gives very good results. Use a good colour quantization algorithm like Neuquant. Another solution is to split the transparency from the image, and encode it as a separate greyscale jpeg.