Nowadays, in order to reduce the negative impacts of inappropriate remarks made by real human streamers (RHS) during live-streaming, and enhance consumer brand forgiveness, brands widely employ human-powered virtual streamers (HPVS) to engage in live-streaming. However, no research has explored the consumer brand forgiveness of RHS and HPVS when they make inappropriate remarks. Based on the meaning transfer theory, a mixed-method approach combining secondary data and situational experiments is used to explore consumer brand forgiveness of the two types of streamers (RHS vs. HPVS) when they make inappropriate remarks, and validate the mediating role of consumer empathy, as well as the moderating role of brand reputation and streamer influence. The results show that consumer brand forgiveness is higher when inappropriate remarks are made by HPVS rather than RHS. Consumer empathy plays a mediating role in the impact path of streamer type (RHS vs. HPVS) on consumer brand forgiveness. Brand reputation and streamer influence moderate the mediating effect of consumer empathy: when brand reputation or streamer influence is high, HPVS (vs. RHS) reduce consumer empathy, which enhances consumer brand forgiveness; when brand reputation or streamer influence is low, HPVS (vs. RHS) do not bring a significant difference in consumer empathy and consumer brand forgiveness.