Modular polyketide synthases (PKSs) have been proposed as promising megasynthases for retrobiosynthesis because of their fitness for rational carbon skeleton design. However, over the last three decades, all engineered PKSs produce carboxylic acids via terminal thioesterase (TE), which significantly narrows available chemical scope. We proposed the possibility of expanding PKS chemical diversity via terminal thioreductase (TR) engineering, and comprehensively characterised PKS TRs as NADPH-dependent enzymes to terminate polyketides with aldehyde. These aldehyde products can be readily converted to industrially valuable alcohols and amines, demonstrated by an engineered rimocidin PKS-TR producing 1,3-butanediol, 1,3-pentanediol, and 1,3-hexanediol with a total titer of 1008 mg/L in Streptomyces albus, or producing 554 mg/L 1-amino-3-alcohols via post-PKS transamination, both in shake flasks. Furthermore, efficient control of the product profile was achieved by coenzyme A (CoA) substrate regulation, as elevating butyryl-CoA level resulted in 1,3-hexanediol product ratio increased from 11% to 77%. We also demonstrated that PKS-TR can produce biosynthetically challenging branched-chain diols, culminating in production of 166 mg/L branched-chain 2-methyl-1,3-diols via acyltransferase exchange.