IP-based media distribution and machine learning enable increasingly personalised media, yet the accessibility needs of autistic and ADHD communities remain largely underserved. We conducted a multi-stage participatory design study (N = 8) with autistic and ADHD participants, combining interviews and co-design activities to identify media disruptors and develop targeted interventions. Through reflexive thematic analysis, we identified key disruptors across sensory, cognitive, and narrative modalities: impeded narrative comprehension, auditory discomfort, dialogue masking in multi-speaker scenes, aversive scene triggers, misaligned pacing, and transient text presentation. From 28 co-designed concepts, we synthesised six functional intervention clusters (spanning intersemiotic translation, audio alteration, content segmentation, and dynamic recaps) and evaluated interactive mockups with a subset of participants (N = 5). Evaluation revealed initial evidence of high perceived utility (Mdn = 3--5), while highlighting that participants required flexible control to adapt interventions to their personal preferences and viewing contexts. This work provides preliminary design insights for integrating user-tailored accessibility for autistic and ADHD audiences into streaming platforms, laying the groundwork for future generalisable evaluation.