Emotion recognition through facial expressions is crucial for interpreting social cues. However, it is often influenced by biases, i.e, systematic recognition advantages for particular emotions. Nevertheless, these biases are inconsistently reported across studies, likely due to methodological variations which underline the necessity for a standardized approach. Traditional face morphing methods can create unnatural-looking stimuli, and may confound the interpretation of emotions. To address this issue, we here introduce STEMorph, a validated stimulus set based on the NimStim set. We employed neutral-anchored morphing and neural-network-generated masks to ensure the natural appearance and integrity of the depicted emotions. We validated our stimulus set by having participants rate each face on a 9-point scale ranging from angry to happy, assessing the perceived emotional intensity. STEMorph's validity was confirmed through linear regression analysis, showing a strong correlation between subjective ratings and targeted emotional states. Moreover, aligning with previous research highlighting gender as a key factor in emotion recognition, STEMorph also exhibits variations in ratings across gender subgroups. STEMorph's reliability was confirmed through a two-week follow-up rating session with a subgroup of the same participants. By introducing a controlled, validated, and ecologically valid stimulus set of emotional faces, STEMorph paves the way for further investigations into facial emotion recognition, deepening our understanding of this vital aspect of human interaction.