Millions of people in the US are eligible to vote despite past criminal convictions, but their voter participation rates are extraordinarily low. In this study, we report the results of a series of randomized controlled trials (RCTs) of mail-based interventions aimed at encouraging people with criminal records to register to vote in North Carolina. We use a novel approach to identify and contact this population, using a combination of administrative data and data from a commercial vendor. In our main experiment, conducted in the fall of 2020, we find that, on average, our mailers increased voter registration by 0.8 percentage points (12%), and voter turnout in the general election by 0.5 percentage points (11%). By contrast, our treatment has no effect on a comparison group of people without criminal records who live in the same neighborhoods. We find suggestive evidence that treatment effects vary across demographic groups and with the content of mailers. For instance, effects were smaller for Black recipients, and smaller when extra "civil rights framing" was added to the mailer text. Overall, we demonstrate that it is possible to identify, contact, and mobilize a marginalized group that is not effectively targeted by existing outreach efforts. Our results speak to how organizations can increase voter registration and turnout among people with criminal records, without necessarily changing laws to broaden eligibility.