diff --git a/examples/offline_inference/voxcpm2/end2end.py b/examples/offline_inference/voxcpm2/end2end.py
index ce404bf962d..687e596018c 100644
--- a/examples/offline_inference/voxcpm2/end2end.py
+++ b/examples/offline_inference/voxcpm2/end2end.py
@@ -74,16 +74,20 @@ def extract_audio(multimodal_output: dict) -> torch.Tensor:
The output processor concatenates per-step delta tensors under
``model_outputs``. Falls back to ``audio`` for backwards compat.
"""
- audio = multimodal_output.get("model_outputs") or multimodal_output.get("audio")
+ audio = multimodal_output.get("model_outputs")
+ if audio is None:
+ audio = multimodal_output.get("audio")
if audio is None:
raise ValueError(f"No audio key in multimodal_output: {list(multimodal_output.keys())}")
if isinstance(audio, list):
- # Take the last valid tensor (most complete audio)
+ # Defensive: usually the output processor consolidates into a single
+ # tensor at request completion, but concatenate here too in case the
+ # caller consumes intermediate (pre-consolidation) outputs.
valid = [torch.as_tensor(a).float().cpu().reshape(-1) for a in audio if a is not None]
if not valid:
raise ValueError("Audio list is empty or all elements are None.")
- return valid[-1]
+ return torch.cat(valid, dim=0) if len(valid) > 1 else valid[0]
return torch.as_tensor(audio).float().cpu().reshape(-1)
diff --git a/examples/online_serving/voxcpm2/gradio_demo.py b/examples/online_serving/voxcpm2/gradio_demo.py
new file mode 100644
index 00000000000..a33a2d9245f
--- /dev/null
+++ b/examples/online_serving/voxcpm2/gradio_demo.py
@@ -0,0 +1,602 @@
+"""Gradio demo for VoxCPM2 TTS with gapless streaming audio playback.
+
+Uses a custom AudioWorklet-based player for gap-free streaming
+(adapted from the Qwen3-TTS demo). Audio is streamed from the vLLM
+server through a same-origin proxy and played via the Web Audio API's
+AudioWorklet, which maintains a FIFO buffer queue and plays samples at
+the audio clock rate.
+
+Usage:
+ # Start the vLLM server first:
+ python -m vllm_omni.entrypoints.openai.api_server \
+ --model openbmb/VoxCPM2 \
+ --stage-configs-path vllm_omni/model_executor/stage_configs/voxcpm2.yaml \
+ --host 0.0.0.0 --port 8000
+
+ # Then launch the demo:
+ python gradio_demo.py --api-base http://localhost:8000
+"""
+
+from __future__ import annotations
+
+import argparse
+import base64
+import io
+import json
+import logging
+
+import gradio as gr
+import httpx
+import numpy as np
+import soundfile as sf
+from fastapi import FastAPI, Request
+from fastapi.responses import Response, StreamingResponse
+
+logger = logging.getLogger(__name__)
+
+SAMPLE_RATE = 48000
+
+# ── AudioWorklet processor (loaded in browser via Blob URL) ──────────
+WORKLET_JS = r"""
+class TTSPlaybackProcessor extends AudioWorkletProcessor {
+ constructor() {
+ super();
+ this.queue = [];
+ this.buf = null;
+ this.pos = 0;
+ this.playing = false;
+ this.played = 0;
+ this.port.onmessage = (e) => {
+ if (e.data && e.data.type === 'clear') {
+ this.queue = []; this.buf = null; this.pos = 0; this.played = 0;
+ if (this.playing) { this.playing = false; this.port.postMessage({type:'stopped'}); }
+ return;
+ }
+ this.queue.push(e.data);
+ };
+ }
+ process(inputs, outputs) {
+ const out = outputs[0][0];
+ for (let i = 0; i < out.length; i++) {
+ if (!this.buf || this.pos >= this.buf.length) {
+ if (this.queue.length > 0) {
+ this.buf = this.queue.shift(); this.pos = 0;
+ } else {
+ for (let j = i; j < out.length; j++) out[j] = 0;
+ if (this.playing) { this.playing = false; this.port.postMessage({type:'stopped', played:this.played}); }
+ return true;
+ }
+ }
+ out[i] = this.buf[this.pos++] / 32768;
+ this.played++;
+ }
+ if (!this.playing) { this.playing = true; this.port.postMessage({type:'started'}); }
+ return true;
+ }
+}
+registerProcessor('tts-playback-processor', TTSPlaybackProcessor);
+"""
+
+PLAYER_HTML = """
+
+"""
+
+
+def _build_player_js() -> str:
+ return f"""
+
+"""
+
+
+def _encode_audio(audio_data: tuple) -> str:
+ sr, audio_np = audio_data
+ if audio_np.dtype in (np.float32, np.float64):
+ audio_np = np.clip(audio_np, -1.0, 1.0)
+ audio_np = (audio_np * 32767).astype(np.int16)
+ elif audio_np.dtype != np.int16:
+ audio_np = audio_np.astype(np.int16)
+ buf = io.BytesIO()
+ sf.write(buf, audio_np, sr, format="WAV")
+ return f"data:audio/wav;base64,{base64.b64encode(buf.getvalue()).decode()}"
+
+
+def create_app(api_base: str):
+ app = FastAPI()
+ _pending: dict[str, dict] = {}
+
+ @app.post("/proxy/v1/audio/speech")
+ async def proxy_speech(request: Request):
+ body = await request.json()
+ req_id = body.get("_req_id")
+ if req_id and req_id in _pending:
+ body = _pending.pop(req_id)
+ logger.info("Proxy: %s", {k: (f"<{len(str(v))} chars>" if k == "ref_audio" else v) for k, v in body.items()})
+ try:
+ client = httpx.AsyncClient(timeout=300)
+ resp = await client.send(
+ client.build_request(
+ "POST",
+ f"{api_base}/v1/audio/speech",
+ json=body,
+ headers={"Authorization": "Bearer EMPTY", "Content-Type": "application/json"},
+ ),
+ stream=True,
+ )
+ except Exception as exc:
+ logger.exception("Proxy connection error")
+ await client.aclose()
+ return Response(content=str(exc), status_code=502)
+ if resp.status_code != 200:
+ content = await resp.aread()
+ await resp.aclose()
+ await client.aclose()
+ return Response(content=content, status_code=resp.status_code)
+
+ async def relay():
+ try:
+ async for chunk in resp.aiter_bytes():
+ yield chunk
+ finally:
+ await resp.aclose()
+ await client.aclose()
+
+ return StreamingResponse(relay(), media_type="application/octet-stream")
+
+ css = """
+ #generate-btn button { width: 100%; }
+ #streaming-player { border: 1px solid var(--border-color-primary) !important; border-radius: var(--block-radius) !important; padding: var(--block-padding) !important; }
+ """
+ theme = gr.themes.Default(
+ primary_hue=gr.themes.Color(
+ c50="#f0f5ff",
+ c100="#dce6f9",
+ c200="#b8cef3",
+ c300="#8eb2eb",
+ c400="#6496e0",
+ c500="#4A90D9",
+ c600="#3a7bc8",
+ c700="#2d66b0",
+ c800="#1f4f8f",
+ c900="#163a6e",
+ c950="#0e2650",
+ ),
+ )
+
+ with gr.Blocks(title="VoxCPM2 TTS Demo") as demo:
+ gr.HTML(f"""
+
+
+
+
VoxCPM2 Streaming Demo
+
+ Served by vLLM-Omni
+ · {api_base}
+ · 48 kHz
+
+
+