# WS08-001: Dedicated Audio Workflow Families (SFX, Voice, Music)

## Scope

Implemented dedicated workflow families for:

- `text-to-audio` (SFX generation)
- `voice-synthesis` (spoken voice generation)
- `music-generation` (music cue composition)

This closes checklist item `WS08-001` by making audio families first-class
across schema, workflow catalog, resolver routing defaults, and taxonomy
metadata.

## Research and implementation alignment

Before implementation, existing patterns were reviewed in:

- Shared workflow catalog and registry metadata (`@isis/workflows`)
- Resolver generation-type inference/default selection (`@isis/generation-api`)
- Existing ComfyUI audio workflow patterns (Stable Audio Open + ACE-Step
  guidance in official Comfy docs)

References:

- [ComfyUI ACE-Step workflow example](https://docs.comfy.org/tutorials/audio/ace-step)
- [ComfyUI Stable Audio Open workflow example](https://docs.comfy.org/tutorials/audio/stable-audio-open)

## Deliverables

1. Schema support for audio workflows in `@isis/workflows`

- Added `audio` workflow category.
- Added `audio` output type/artifact type.
- Added audio output formats (`wav`, `mp3`, `flac`, `ogg`, `m4a`).
- Added `audio` workflow directory to loader category discovery.

2. Dedicated workflow families added to catalog

- `sfx-sound-designer`
- `voice-performance-synthesis`
- `music-cue-composer`

Each workflow includes canonical metadata, inputs, node graph stubs, prompt
construction, and `output_spec` with `audio` package contracts.

3. Resolver and routing updates in `@isis/generation-api`

- Added default workflow mappings:
  - `text-to-audio -> sfx-sound-designer`
  - `voice-synthesis -> voice-performance-synthesis`
  - `music-generation -> music-cue-composer`
- Extended catalog inference for `outputType: "audio"` with deterministic
  fingerprinting rules.
- Extended output schema conversion to map `audio` output types correctly.
- Added MIME mappings for common audio formats.

4. Taxonomy + metadata updates

- Regenerated workflow taxonomy registry and markdown map with 48 workflow
  mappings.
- Added audio category defaults and audio workflow overrides in taxonomy
  generator.

5. Test coverage updates

- Updated registry/schema tests for new audio category/type/format support.
- Added validation test for audio workflow file loading.
- Updated resolver tests for new audio-family deterministic resolution.
- Kept unresolved-path guard coverage by switching unresolved fixtures to
  `inpainting` (still unmapped by design).

## Verification

- `pnpm --filter @isis/workflows exec vitest run src/workflows.spec.ts --reporter=dot`
- `pnpm --filter @isis/generation-api exec vitest run src/workflows/executor.spec.ts --reporter=dot`
- `node scripts/isis/validate-v2-workflow-taxonomy-registry.mjs`
- `node scripts/isis/validate-v2-compliance-gate.mjs`

## Result

`WS08-001` is now implemented with dedicated audio workflow families wired
through canonical ISIS workflow contracts and runtime resolution.
