Objective To observe the efficacy of respiratory rehabilitation in patients with coronavirus disease 2019 (COVID-19) and the effect of different respiratory training methods. Methods Twenty-one patients of common or severe COVID-19 with stable condition and positive rehabilitation willingness were selected from a ward of Wuhan Jinyintan Hospital during Feb. 4 to Mar. 25, 2020. They were evenly randomized into 3 groups. In addition to the conventional drugs and symptomatic and supportive treatments, the 3 groups of patients were given Shunqing respiratory trainer (group A), Acapella respiratory trainer (group B) and Leventon Spiro-Ball respiratory trainer (group C), respectively, for respiratory function and expectoration trainings, twice a day, 30 min each time, for 7 days. Four patients with stable condition but without positive respiratory rehabilitation willingness were selected as the control group, and they were only given conventional drugs and symptomatic and supportive treatments. Modified Borg's rating of perceived exertion (RPE) score, respiratory rate, pulse oxygen saturation (SpO₂) without oxygen support, total sputum volume, and activities of daily living (ADL) score were assessed before and 7 days after rehabilitation training. Results No significant differences were observed in demographic characteristics (such as age, gender), baseline modified RPE score, SpO₂ value without oxygen support, constituent ratio of sputum volume or ADL score among the 4 groups before training (all P>0.05), but there was significant difference in respiratory rate (P=0.040). After training, the modified RPE scores, respiratory rates, SpO₂ values without oxygen support and ADL scores in groups A, B and C were significantly improved compared with those before training (all P<0.05); the modified RPE scores, respiratory rates, SpO₂ values without oxygen support and ADL scores in groups A and B were significantly different from those in the control group (all P<0.05), while only respiratory rate in group C was significantly different from that in the control group (P<0.05); there were no significant differences in the above 4 indexes between groups A, B and C (all P>0.05). The post-training sputum volume of the 4 groups were decreased compared with that pre-training, and there was significant difference in the constituent ratio of sputum volume of groups A, B and C before and after training (P<0.05), while there was no significant difference in the control group before and after training (P>0.05). Conclusion Respiratory rehabilitation assisted by respiratory trainers can effectively reduce sputum secretion, improve dyspnea symptoms, respiratory function and activities of daily living in patients with common or severe COVID-19. There are no significant differences in the effects of different respiratory trainers. Further large sample multicenter studies are needed.