Introduction: The embryonic thermal programming (TM) in birds has been shown to impact several physiological parameters such as resistance to thermal stress, muscle growth or immunity. In mule ducks, it has recently been shown that TM can induce metabolic programming resulting in increased liver weight and fat storage after overfeeding. However, a decrease in hatchability and foie gras quality was also observed, suggesting that this technique needs to be optimized. Here, we tested a new thermal manipulation condition determined with the objective of avoiding negative impacts while maintaining or improving liver properties. Methods: The eggs of the control group were incubated at 37.6°C during the whole incubation period while those of the experimental group (TM group) were incubated at 39.3°C 16 h/24 h from the 11 th day of incubation to the 21 st . After hatching, all the animals were fed and raised under the same conditions until the age of 12 weeks. At this stage, one part of the animals was overfed and then slaughtered 2 h (to measure rapid changes in metabolism) or 10 h after the last meal (to obtain the best technological yields), while the other part was ration-fed and slaughtered 2 h after the last meal, at the same age. Results: An 8% increase in foie gras production was measured in the TM group compared to the control group without altering the quality of the final product (nor hatchability), confirming the successful optimization of the metabolic programming. Interestingly, these results allowed us not to reject the previously suggested hypothesis of a potential delay in metabolic processes involved in liver fattening in programmed animals, in particular by measuring a trend reversal regarding the amount of total hepatic lipids in both groups at 2 h and then 10 h after the last meal. Discussion: This study therefore validates the optimization of metabolic programming by embryonic thermal manipulation for duck liver fattening. The understanding of the mechanisms of embryonic thermal programming in birds remains today very incomplete and the search for epigenetic marks (main hypothesis of the concept of programming) at the origin of the observed phenotypes could be the next step of this work.