This paper deals with small-power energy harvesting from heat. It can be achieved using both thermoelectric and pyroelectric effects. In the first case, temperature gradients are necessary. The main difficulty of thermoelectric energy harvesting is imposing a large temperature gradient. This requires huge heat flows because of the limited surface heat exchanges and the large heat conductivity of thermoelectric materials. This results in a drastic decrease of power and the efficiency of conversion. In case of pyroelectric energy harvesting, a time varying temperature is necessary. Although such a temperature time profile is hard to find, the overall optimization is easier than the thermoelectric strategy. Indeed, it depends much less on heat exchange between the sample and the outer medium, than on heat capacity that dimensions optimization may easily compensate. As a consequence, it is shown that the efficiency and output power may be much larger using pyroelectric energy harvesting than thermoelectric methods. For instance, using a limited temperature gradient due to the limited heat exchange, a maximum efficiency of 1.7% of Carnot efficiency can be expected using a thermoelectric module. On the contrary, a pyroelectric device may reach an efficiency up to 50% of Carnot efficiency. Finally, an illustration shows an estimation of the output power that could be expected from natural time variations of temperature of a wearable device. Power peaks up to 0.2 mW cm⁻³ were found and a mean power of 1 µW cm⁻³ on average was determined within 24 h testing.