Long-lasting chemiluminescence (CL) emissions are necessary for improving the detection accuracy and expanding the application scope. Here, we have synthesized three oil-in-water (O/W) multicolor protein capsules (LCBA, F/LCBA, and RB/F/LCBA) using a simple ultrasound method and have engineered specific target-triggered catalytic hairpin assembly on their surface and chemiluminescence resonance energy transfer inside. Consequently, three multicolor capsules exhibit excellent structural stability, generate blue-, green-, and red-colored emissions when reacting with H2O2, have long-lasting CL emission over 1 h, and successfully achieve the accurate multiple visualization detection of avian influenza virus subtype targets. Without the need for complex instruments and analysis procedures, the CL imaging assays can be carried out and recorded with a common smartphone. The detection limits for visualizing H1N1, H7N9, and H5N1 are 5.5, 7.6, and 9.0 pM, respectively. There is a linear range between 20.0 and 625 pM and excellent selectivity against interfering DNA. Furthermore, visualization detection has been successfully applied for the detection of H1N1, H7N9, and H5N1 in healthy human serum samples. With these merits, this facile, ultrasensitive, and multiple visualization sensor has potential applications in point-of-care testing and early diagnosis.