README.md

---
title: Forge-Audio — Music, SFX & Qwen3-TTS Voice
emoji: 🔊
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 6.18.0
app_file: app.py
pinned: true
license: apache-2.0
short_description: Music, SFX and Qwen3-TTS voice (design/gen/clone) for games.
---

# Forge-Audio

**Music + SFX + Full Qwen3-TTS Voice Pipeline** for game assets.

Strong emphasis on the **Qwen3-TTS** capabilities you requested:
- Voice design from natural language character descriptions
- Preset voice generation
- Recorded voice cloning + line generation

Part of the open **Forge Bricks** collection (9 composable Gradio Spaces) for a complete game asset creation system.

## Capabilities
- `/design_voice` — Character description → voice profile (Qwen primary)
- `/generate_voice` — Text + voice description → spoken dialogue
- Music generation endpoint (mood/genre → loopable tracks)
- SFX generation (description + trigger context)
- Rich manifests with lineage for the rest of the pipeline

## Quick Start (Local)

```bash
./install.sh   # from forge-bricks root
source .venv/bin/activate
cd forge-audio
python app.py
```

## HF Space
- Designed for **ZeroGPU** (voice operations are relatively light).
- Use the provided `requirements.txt`.

## Daggr / Agent Integration
```python
from daggr import GradioNode
voice = GradioNode("jkorstad/forge-audio", api_name="/design_voice", ...)
line  = GradioNode("jkorstad/forge-audio", api_name="/generate_voice", ...)
```

## License Notes
Qwen3-TTS design/generation is often non-commercial. The brick surfaces `commercial_ok` per output and provides Kokoro fallback paths where appropriate. Check the per-generation `manifest.json`.

## In the Pipeline
Use after character concepting (forge-visuals / forge-3d) to give your NPCs and creatures voices, then combine with rig/anim bricks.

See the main Forge Bricks README and approved plan for full composition examples with daggr and ForgeDNA.

*Open building blocks for creators and agents.*