Carbon and nitrogen stable isotope technology is an important tool for delineating the nutritional positions and structure of ecosystem food webs. Muscles are commonly used to determine stable isotopes, but changes in the muscle lipid contents affect the results. To investigate this effect, this study selected 45 samples of from the offshore waters of southern Zhejiang and divided them into two groups, 3.5< C:N ≤ 4.0 (group 1) and C:N >4.0 (group 2). Changes in the carbon-nitrogen stable isotope ratios before and after lipid removal and the best lipid correction model were selected from multiple regression models. Pearson correlation analysis revealed significant positive correlations of δ¹³C (stable isotope of carbon), δ¹⁵N (stable isotope of nitrogen), and C:N before and after lipid removal in group 1 (P<0.05). δ¹⁵N before and after lipid removal in group 2 was significantly positively correlated (<0.05), but C:N and Δδ¹³C were positively correlated (<0.05). According to AIC (Akaike information criterion), the linear model has the best fitting degree for the samples with C:N in the range of 3.5 to 4.0, and the fitting equation is Δδ¹³C=1.30×(C:N)-3.69. The nonlinear model has the highest fitting degree for samples with C:N more than 4.0, and the fitting equation is Δδ¹³C=-5.71+5.12×ln(C:N). This study found that the lipid contents significantly affect the determination of carbon stable isotopes. Pretreatment by lipid removal is required for isotopic analysis, especially for muscle samples with high lipid contents. The lipid normalization model simplifies sample processing and has wide-ranging applicability. It is important to select the correct model for different species in diverse locations with varying fat contents.