Neuroprostheses have been used clinically for decades, to help restore or preserve brain functions, when pharmaceutical treatments are inefficient. Although great progress in the field has been made over the years to interface with the nervous system, surface neural interfaces could still benefit from improvement for efficient, selective, and long-term functional use. Clinical surface neural interfaces are traditionally made of platinum disks embedded in a thick silicone matrix, making them relatively rigid and bulky, while on the contrary, nervous tissues are soft. Hence, recent trends in bioelectronics aim at developing "soft" neural interfaces. They are made both flexible and stretchable, to better conform to the various topographies of the nervous system, in particular the brain, and to accommodate for micro and macro movements. In this thesis, the aim was to evaluate the suitability of soft surface electrodes to contribute to neuroprosthetic advancements. Specifically, we aimed at validating soft implants made of a thin silicone matrix as substrate and encapsulation material that embeds thin conducting metal layers, either micro-cracked gold or micropatterned platinum. Using a novel technology to develop neuroprosthetic strategies requires a full understanding of its usability, with robust animal models and defined read-outs for testing. First, we investigated the opportunity of soft neural implants to improve rehabilitation in profound deafness. We addressed the limitations of clinically available Auditory Brainstem Implant (ABI), by developing a soft ABI that was evaluated electrophysiologically and behaviorally in vivo in non-human primate (NHP) models. By stimulating the Cochlear Nucleus (CN), we were able to generate auditory-like responses at the cortex that could drive behavior. In addition, we proposed a novel approach to specifically activate the auditory pathway for hearing restoration, by targeting the auditory cortex directly, using a soft electrocorticography (ECoG) array (Auditory Cortex Implant (ACI)). This strategy, offering a less invasive and higher resolution alternative compared to the ABI, was validated both in rodents and NHP. Notably, auditory cortex driven responses could be modulated to resemble frequency-specific cortical activation in the rat, and auditory cortex stimulation could induce behaviorally relevant cues in macaques. Next, we delved into the field of traumatic brain injury and particularly secondary insults linked to spreading depolarizations (SDs). Highlighting the importance of neuromonitoring and treatment development, we have tested soft ECoG arrays' ability to reliably record SDs in different animal models (rat and pig) and established a closed-loop neuroprosthetic platform for modulation of SD waves in a rat model, yielding preliminary encouraging results. Altogether, by setting up meticulous analysis pipelines for cortical and behavioral data, investigating stimulation strategies and optimizing implant designs, promising results regarding the use of soft neural implants could be demonstrated in this thesis. In particular, we showed favorable outcomes towards restoring hearing sensation via the brainstem or cortex, as well as improvements of clinical practices for monitoring SDs and providing a platform for treatment evaluation. Both current and future neurotherapies in the clinic could benefit from the use of improved contact to the host tissue using soft bioelectronic interfaces.