This paper explores the potential of AI text-to-image diffusion models (e.g. DALLE-2 and Midjourney) to support the early phase design of new digital musical instruments in collaboration with Disabled musicians. The paper presents initial findings from two speculative design workshops attended by Disabled participants who are affiliated with the London-based inclusive arts organisation Joy of Sound. The workshops included activities enabling participants to co-create speculative images of new instruments, drawing on their contributions. These included the overall appearance of the instrument, constituent materials and other design characteristics. The paper discusses the generated images and examines how diffusion models can be a useful tool to support the conceptual co-design phase of bespoke accessible instruments. The project findings indicate that diffusion models can be useful as a facilitatory tool for idea generation in the initial stages of bespoke instrument design.