|
|
<!DOCTYPE html> |
|
|
<html> |
|
|
<head> |
|
|
<meta charset="utf-8"> |
|
|
<title>MagentaRT Research API</title> |
|
|
<style> |
|
|
body { font-family: Arial, sans-serif; max-width: 860px; margin: 48px auto; padding: 0 20px; color:#111; } |
|
|
code, pre { background:#f6f8fa; border:1px solid #eaecef; border-radius:6px; padding:2px 6px; } |
|
|
pre { padding:12px; overflow:auto; } |
|
|
.muted { color:#555; } |
|
|
ul { line-height: 1.8; } |
|
|
</style> |
|
|
</head> |
|
|
<body> |
|
|
<h1>🎵 MagentaRT Research API</h1> |
|
|
<p class="muted"><strong>Purpose:</strong> AI music generation for iOS/web app research using Google's MagentaRT.</p> |
|
|
|
|
|
<h2>Available Endpoints</h2> |
|
|
<ul> |
|
|
<li><code>POST /generate</code> – Generate 4–8 bars of music (HTTP, bar-aligned)</li> |
|
|
<li><code>POST /jam/start</code> – Start continuous jamming (HTTP)</li> |
|
|
<li><code>GET /jam/next</code> – Get next chunk (HTTP)</li> |
|
|
<li><code>POST /jam/consume</code> – Confirm a chunk as consumed (HTTP)</li> |
|
|
<li><code>POST /jam/stop</code> – End session (HTTP)</li> |
|
|
<li><code>WEBSOCKET /ws/jam</code> – Realtime streaming (<code>mode="rt"</code>)</li> |
|
|
<li><code>GET /docs</code> – API documentation (Gradio)</li> |
|
|
</ul> |
|
|
|
|
|
<h2>WebSocket Quick Start (rt mode)</h2> |
|
|
<p>Connect to <code>wss://<your-space>/ws/jam</code> and send:</p> |
|
|
<pre>{ |
|
|
"type": "start", |
|
|
"mode": "rt", |
|
|
"binary_audio": false, |
|
|
"params": { |
|
|
"styles": "warmup", |
|
|
"temperature": 1.1, |
|
|
"topk": 40, |
|
|
"guidance_weight": 1.1, |
|
|
"pace": "realtime", // or "asap" to bootstrap quickly |
|
|
"max_decode_frames": 50 // default ~2.0s; try 36–45 on smaller GPUs |
|
|
} |
|
|
}</pre> |
|
|
<p>Update parameters live:</p> |
|
|
<pre>{ |
|
|
"type": "update", |
|
|
"styles": "jazz, hiphop", |
|
|
"style_weights": "1.0,0.8", |
|
|
"temperature": 1.2, |
|
|
"topk": 64, |
|
|
"guidance_weight": 1.0, |
|
|
"pace": "realtime", |
|
|
"max_decode_frames": 40 |
|
|
}</pre> |
|
|
<p>Stop:</p> |
|
|
<pre>{"type":"stop"}</pre> |
|
|
|
|
|
<h2>Notes</h2> |
|
|
<ul> |
|
|
<li>Audio: 48 kHz stereo, ~2.0 s chunks by default with ~40 ms crossfade.</li> |
|
|
<li>L40S 48GB: faster than realtime → prefer <code>pace: "realtime"</code>.</li> |
|
|
<li>L4 24GB: slightly under realtime even with pre-roll and tuning.</li> |
|
|
<li>For sustained realtime, target ~40 GB VRAM per active stream (e.g., A100 40GB or ≈35–40 GB MIG slice).</li> |
|
|
</ul> |
|
|
|
|
|
<p class="muted"><strong>Licensing:</strong> Uses MagentaRT (Apache 2.0 + CC-BY 4.0). Users are responsible for outputs.</p> |
|
|
<p>See <a href="/docs">/docs</a> for full API details and client examples.</p> |
|
|
</body> |
|
|
</html> |
