Many Americans save too little, leaving them vulnerable to unexpected financial shocks. Finding ways to help Americans develop emergency savings funds could greatly improve welfare. A wealth of previous literature has demonstrated the central roles played by patience and self-control in achieving sufficient savings. When people lack patience or self-control, well-designed interventions may help improve financial stability. Increasingly, interventions intended to improve savings behavior have taken the form of externally restricted accounts such as 'commitment accounts' that include hefty fees for early withdrawal or that disallow withdrawals altogether for a pre-specified time. Yet, such hard commitment accounts may not appeal to impatient individuals, those who do not anticipate their own self-control problems, or to the poor for whom restrictions on scarce funds can be particularly painful. We test a new 'soft' commitment account that asks borrowers to think about their savings goals, how it would feel to achieve them, and make a pledge to work towards these goals (potentially increasing one's intrinsic motivation), yet has no external restrictions on savings behavior. In a six-month randomized savings experiment we find that such soft commitments can significantly increase amounts saved on day one relative to either a hard commitment account (with external restrictions on withdrawals) or a traditional savings account. Additionally, the soft commitments significantly increased final savings balances relative to no form of commitment and were particularly effective for impatient individuals. However, despite the inherent illiquidity, the hard commitment account proved most effective in building savings balances amongst our participants at the end of six months.