Can a technically synthesized persona function as a credible and effective communicator? Deep synthesis personas are showing distinctive potential in science communication, yet their inherently synthetic nature can trigger psychological resistance and raise doubts about message credibility. This study examines how shortening the psychological distance between audiences and deep synthesis personas can improve communication effects. We adopt the Elaboration Likelihood Model (ELM) as the primary framework and incorporate psychological distance theory to integrate the analysis. Using a sample of deep synthesis persona science videos from Bilibili (N=193), we combined web scraping with manual coding and estimated regression models to test how features related to content and technology shaping effects, including reach, engagement, and endorsement. The results show that, on the content route, multi-character interactive discussion formats, as compared with solo monologue formats, and narration delivered through reanimated authentic historical figures, as compared with fictional characters, effectively reduce psychological distance and foster a more active discussion climate. On the technology route, higher visual realism and smoother, more natural facial animation increase viewers’ sense of presence and perceived authenticity, shorten psychological distance from the persona, and in turn raise interactive behaviors and positive feedback. In parallel, peripheral cues such as creator prominence exert significant positive effects on diffusion metrics, underscoring the importance of source credibility. These findings offer practical guidance for video production and dissemination strategy and provide a theoretical reference for understanding the opportunities and social implications of deploying deep synthesis technologies in science communication.