Photoisomerization
The most direct way to accomplish this is through photoisomerization. Molecules like azobenzene(shown below) can be switched between two different configurations using different wavelengths of light.
If we arrange our photoisomers in a polymer just right, we can obtain light induced deformation at the macroscale One class of polymers that does this are liquid crystal elastomers. It is also worth noting that the initial steps our phototransduction in our eyes rely on retinal MECHANICALLY changing the shape of the protein opsin. Although making a liquid crystal elastomer which can operate at acoustic frequencies would probably be difficult.
Photochemical reactions
A fun way to convert an optical signal into mechanical motion is to use light triggered chemical reactions. Hydrogen and chlorine will react with each other explosively when exposed to UV light. It is not inconceivable that a carefully designed reaction chamber filled with flowing hydrogen and chlorine could be made to explode at acoustic frequencies.
Photoacoustics
There's also the photoacoustic effect, where an oscillating light signal can be used to make a fluid heat up and cool down sufficiently fast so as to make sound. Such sound is typically quite soft unless one uses a very strong optical signal. However, fluidics enables us to amplify these small acoustic signals into something much louder or capable of controlling pneumatic or hydraulic machinery. The below image taken from here illustrates how this is possible.
Light is focused on a fluidic amplifer causing fluid to expand and displace a fluid jet result in said fluid jet moving more fluid to one port than another resulting in amplification of the small expansion of fluid.
Fluidics has been used to amplify photoacoustic signals from a 0.7 mW light source to drive a control valve for a gas chromatograph. In addition, fluidics has also been used to amplify photoacoustic signals to drive a pneumatic robot arm. Perhaps the most bizarre use of using fluidics to amplify optical signals was to make a completely nonelectric remote control system for an ICBM interceptor. An audio receiver based on this principle has been patented too, I can't find any information that it was actually built, but there is good reason to believe that it's possible. The aforementioned missile guidance system's preamplifier stage was able to amplify signals up to 5 KHz, which is sufficient for speech transmission.