The model supports 83 languages through one API endpoint and can be used by setting the Fish API model header to s2.1-pro-free. Fish Audio describes S2.1 Pro as its current state-of-the-art voice model, with roughly 90 ms time to first audio on standard API calls, higher concurrency throughput, and the same endpoint structure as paid plans.
S2.1 Pro is useful for voice agents, audiobook and narration pipelines, game NPC dialogue, multilingual product experiences, and prototype voice-cloning workflows. The free access is governed by fair-use constraints and does not include production SLA guarantees, so teams can evaluate quality and latency before moving commercial workloads to paid terms.
