BACKGROUND:In low- and middle-income countries, mental health training often includes sending few generalist clinicians to specialist-led programs for several weeks. Our objective is to develop and test a video-assisted training model addressing the shortcomings of traditional programs that affect scalability: failing to train all clinicians, disrupting clinical services, and depending on specialists. METHODS:We implemented the program -video lectures and on-site skills training- for all clinicians at a rural Nepali hospital. We used Wilcoxon signed-rank tests to evaluate pre- and post-test change in knowledge (diagnostic criteria, differential diagnosis, and appropriate treatment). We used a series of 'Yes' or 'No' questions to assess attitudes about mental illness, and utilized exact McNemar's test to analyze the proportions of participants who held a specific belief before and after the training. We assessed acceptability and feasibility through key informant interviews and structured feedback. RESULTS:For each topic except depression, there was a statistically significant increase (Δ) in median scores on knowledge questionnaires: Acute Stress Reaction (Δ = 20, p = 0.03), Depression (Δ = 11, p = 0.12), Grief (Δ = 40, p < 0.01), Psychosis (Δ = 22, p = 0.01), and post-traumatic stress disorder (Δ = 20, p = 0.01). The training received high ratings; key informants shared examples and views about the training's positive impact and complementary nature of the program's components. CONCLUSION:Video lectures and on-site skills training can address the limitations of a conventional training model while being acceptable, feasible, and impactful toward improving knowledge and attitudes of the participants.