skald-core::narrate ships a thin reqwest client + voice DB access (get_by_name, get_default, get_by_id). The boundary is the f5-tts container's purpose-built FastAPI sidecar (python lives there because torch + transformers + safetensors do); skald never touches python. CLI: skald narrate --chapter <uuid> [--voice slug] [--speed 1.0]. Voice resolution: --voice flag → story.preferred_voice_id → system default. Persists narration_runs row (engine='f5-tts', engine_version pinned, status: running → succeeded|failed). Output path stored is the f5-tts container's view (/audio/<story>-<n>-<run>.wav); web playback wiring deferred. |
||
|---|---|---|
| .. | ||
| src | ||
| Cargo.toml | ||