fix: use html audio to load stream

src/components/audio-animation/index.tsx

@@ -189,4 +189,4 @@ const AudioAnimation: React.FC<AudioAnimationProps> = (props) => {
   );
 };
 
-export default React.memo(AudioAnimation);
+export default AudioAnimation;

src/components/audio-player/raw-audio-player.tsx

@@ -18,8 +18,9 @@ const RawAudioPlayer: React.FC<AudioPlayerProps> = forwardRef((props, ref) => {
   // ========================================================
 
   const initAudioContext = useCallback(() => {
-    audioContext.current = new (window.AudioContext ||
-      window.webkitAudioContext)();
+    audioContext.current = new (
+      window.AudioContext || (window as any).webkitAudioContext
+    )();
 
     analyser.current = audioContext.current.createAnalyser();
     analyser.current.fftSize = 512;
@@ -52,7 +53,9 @@ const RawAudioPlayer: React.FC<AudioPlayerProps> = forwardRef((props, ref) => {
     });
 
     audioRef.current.addEventListener('timeupdate', () => {
+      const current = audioRef.current?.currentTime || 0;
       props.onTimeUpdate?.();
+      props.onAudioProcess?.(current);
     });
 
     audioRef.current.addEventListener('ended', () => {
@@ -85,11 +88,6 @@ const RawAudioPlayer: React.FC<AudioPlayerProps> = forwardRef((props, ref) => {
       props.onVolumeChange?.();
     });
 
-    audioRef.current.addEventListener('audioprocess', () => {
-      const current = audioRef.current?.currentTime || 0;
-      props.onAudioProcess?.(current);
-    });
-
     audioRef.current.addEventListener('playing', () => {
       props.onPlaying?.();
     });
@@ -111,16 +109,44 @@ const RawAudioPlayer: React.FC<AudioPlayerProps> = forwardRef((props, ref) => {
 
   useImperativeHandle(ref, () => ({
     play: () => {
-      audioRef.current?.play();
+      if (audioRef.current) {
+        // If playback has ended, reset to beginning
+        if (audioRef.current.currentTime >= audioRef.current.duration) {
+          audioRef.current.currentTime = 0;
+        }
+        audioRef.current.play();
+      }
     },
     pause: () => {
       audioRef.current?.pause();
+    },
+    seekTo: (ratio: number) => {
+      if (audioRef.current && audioRef.current.duration) {
+        audioRef.current.currentTime = ratio * audioRef.current.duration;
+      }
+    },
+    // Compatibility layer for wavesurfer interface
+    wavesurfer: {
+      current: {
+        play: () => {
+          if (audioRef.current) {
+            // If playback has ended, reset to beginning
+            if (audioRef.current.currentTime >= audioRef.current.duration) {
+              audioRef.current.currentTime = 0;
+            }
+            return audioRef.current.play();
+          }
+          return Promise.resolve();
+        },
+        isPlaying: () => {
+          return audioRef.current ? !audioRef.current?.paused : false;
+        }
+      }
     }
   }));
 
   useEffect(() => {
     if (audioRef.current) {
-      console.log('audioRef.current', audioRef.current, props.url);
       initEnvents();
     }
     return () => {
@@ -138,7 +164,6 @@ const RawAudioPlayer: React.FC<AudioPlayerProps> = forwardRef((props, ref) => {
       audioRef.current?.removeEventListener('seeked', () => {});
       audioRef.current?.removeEventListener('seeking', () => {});
       audioRef.current?.removeEventListener('volumechange', () => {});
-      audioRef.current?.removeEventListener('audioprocess', () => {});
       audioRef.current?.removeEventListener('playing', () => {});
       audioRef.current?.removeEventListener('loadedmetadata', () => {});
       audioRef.current?.removeEventListener('ended', () => {});
@@ -146,6 +171,13 @@ const RawAudioPlayer: React.FC<AudioPlayerProps> = forwardRef((props, ref) => {
     };
   }, [audioRef.current]);
 
+  // Reload audio when URL changes
+  useEffect(() => {
+    if (audioRef.current && props.url) {
+      audioRef.current.load();
+    }
+  }, [props.url]);
+
   return (
     <audio
       controls

src/components/speech-content/audio-player.tsx

@@ -36,8 +36,9 @@ const AudioPlayer: React.FC<
   const mediaElement = useRef<any>(null);
 
   const initAudioContext = useCallback(() => {
-    audioContext.current = new (window.AudioContext ||
-      window.webkitAudioContext)();
+    audioContext.current = new (
+      window.AudioContext || window.webkitAudioContext
+    )();
 
     analyser.current = audioContext.current.createAnalyser();
     analyser.current.fftSize = 512;
@@ -218,4 +219,4 @@ const AudioPlayer: React.FC<
   );
 });
 
-export default React.memo(AudioPlayer);
+export default AudioPlayer;

src/components/speech-content/index.tsx

@@ -4,16 +4,34 @@ import SpeechItem from './speech-item';
 interface SpeechContentProps {
   dataList: any[];
   loading?: boolean;
+  onPlay?: () => void;
+  onPause?: () => void;
+  playerRef?: React.RefObject<any>;
+  isPlaying?: boolean;
+  isStream?: boolean;
+  analyserData?: {
+    data: Uint8Array;
+    analyser: any;
+  };
 }
 
 const SpeechContent: React.FC<SpeechContentProps> = (props) => {
   return (
-    <>
+    <div>
       {props.dataList.map((item) => (
-        <SpeechItem key={item.uid} {...item} />
+        <SpeechItem
+          key={item.uid}
+          {...item}
+          isStream={props.isStream}
+          isPlaying={props.isPlaying}
+          onPlay={props.onPlay}
+          onPause={props.onPause}
+          playerRef={props.playerRef}
+          analyserData={props.analyserData}
+        />
       ))}
-    </>
+    </div>
   );
 };
 
-export default React.memo(SpeechContent);
+export default SpeechContent;

src/components/speech-content/speech-item.tsx

@@ -8,8 +8,14 @@ import { useIntl } from '@umijs/max';
 import { Button, Slider, Tooltip } from 'antd';
 import dayjs from 'dayjs';
 import _, { throttle } from 'lodash';
-import React, { useCallback, useRef, useState } from 'react';
-import AudioPlayer from './audio-player';
+import React, {
+  useCallback,
+  useEffect,
+  useMemo,
+  useRef,
+  useState
+} from 'react';
+import RawAudioPlayer from '../audio-player/raw-audio-player';
 import './styles/index.less';
 import './styles/slider-progress.less';
 
@@ -34,10 +40,25 @@ interface SpeechContentProps {
   format: string;
   speed: number;
   audioUrl: string;
+  onPlay?: () => void;
+  onPause?: () => void;
+  isStream?: boolean;
+  isPlaying?: boolean;
+  playerRef?: React.RefObject<any>;
+  analyserData?: {
+    data: Uint8Array;
+    analyser: any;
+  };
 }
 const SpeechItem: React.FC<SpeechContentProps> = (props) => {
+  const { isStream } = props;
+  console.log(
+    'Rendering SpeechItem with props:',
+    props.isStream,
+    props.analyserData
+  );
   const intl = useIntl();
-  const [isPlay, setIsPlay] = useState(props.autoplay);
+  const [isPlay, setIsPlay] = useState(props.autoplay || props.isPlaying);
   const [duration, setDuration] = useState<number>(0);
   const [animationSize, setAnimationSize] = useState({ width: 900, height: 0 });
   const [currentTime, setCurrentTime] = useState(0);
@@ -48,23 +69,47 @@ const SpeechItem: React.FC<SpeechContentProps> = (props) => {
   const wrapper = useRef<any>(null);
   const ref = useRef<any>(null);
 
-  const handlePlay = useCallback(async () => {
+  useEffect(() => {
+    setIsPlay(props.autoplay || props.isPlaying);
+  }, [props.autoplay, props.isPlaying]);
+
+  const isPCMStream = useMemo(() => {
+    return props.audioUrl?.startsWith('pcm-stream://');
+  }, [props.audioUrl]);
+
+  // Sync internal ref with external playerRef if provided
+  React.useEffect(() => {
+    if (props.playerRef) {
+      (props.playerRef as any).current = ref.current;
+    }
+  }, [props.playerRef, ref.current]);
+
+  const onPause = () => {
+    ref.current?.pause();
+    props.onPause?.();
+  };
+
+  const onPlay = async () => {
+    await ref.current?.wavesurfer.current?.play();
+    props.onPlay?.();
+  };
+
+  const handlePlay = async () => {
     try {
       if (ref.current?.wavesurfer.current?.isPlaying()) {
-        ref.current?.pause();
+        onPause();
         setIsPlay(false);
         return;
       } else {
-        await ref.current?.wavesurfer.current?.play();
+        await onPlay();
         setIsPlay(true);
       }
     } catch (error) {
       console.log('error:', error);
     }
-  }, [ref.current]);
+  };
 
   const handleOnAnalyse = useCallback((data: any, analyser: any) => {
-    console.log('data+++++++++++++++++=:', data);
     setAudioChunks((pre: any) => {
       return {
         data: data,
@@ -76,9 +121,11 @@ const SpeechItem: React.FC<SpeechContentProps> = (props) => {
   const handleOnFinish = useCallback(() => {
     setIsPlay(false);
   }, []);
+
   const handleOnPlay = useCallback(() => {
     setIsPlay(true);
   }, []);
+
   const handleOnPause = useCallback(() => {
     setIsPlay(false);
   }, []);
@@ -87,22 +134,9 @@ const SpeechItem: React.FC<SpeechContentProps> = (props) => {
     setCurrentTime(current);
   }, 100);
 
-  const handleOnAudioprocess = useCallback(
-    (current: number) => {
-      console.log('current:', current);
-      throttleUpdateCurrentTime(current);
-    },
-    [throttleUpdateCurrentTime]
-  );
-
-  const handleReay = useCallback((duration: number) => {
-    setIsPlay(props.autoplay);
-    setDuration(duration);
-  }, []);
-
-  const handleOnClick = useCallback((value: number) => {
-    console.log('current:', value);
-  }, []);
+  const handleOnAudioprocess = (current: number) => {
+    throttleUpdateCurrentTime(current);
+  };
 
   const handleAnimationResize = useCallback((size: any) => {
     setAnimationSize({
@@ -121,7 +155,6 @@ const SpeechItem: React.FC<SpeechContentProps> = (props) => {
   };
 
   const handleReady = useCallback((duration: number) => {
-    console.log('ready+++++++++++++++', duration);
     setDuration(duration);
   }, []);
 
@@ -130,9 +163,16 @@ const SpeechItem: React.FC<SpeechContentProps> = (props) => {
     setCurrentTime(value);
   }, []);
 
+  const convertFormat = () => {
+    if (props.format === 'pcm') {
+      return 'wav';
+    }
+    return props.format;
+  };
+
   const onDownload = useCallback(() => {
     const url = props.audioUrl || '';
-    const filename = `audio-${dayjs().format('YYYYMMDDHHmmss')}.${props.format}`;
+    const filename = `audio-${dayjs().format('YYYYMMDDHHmmss')}.${convertFormat()}`;
 
     const link = document.createElement('a');
     link.href = url;
@@ -142,6 +182,71 @@ const SpeechItem: React.FC<SpeechContentProps> = (props) => {
     link.remove();
   }, [props.audioUrl, props.format]);
 
+  console.log('isPCMStream:', isPCMStream, isPlay);
+
+  console.log('props.analyserData:', props.analyserData);
+
+  const renderPlayerActions = () => {
+    if (isPCMStream) {
+      return null;
+    }
+    return (
+      <div>
+        <Slider
+          className="slider-progress"
+          value={currentTime}
+          max={duration}
+          step={0.01}
+          onChange={handleSliderChange}
+        ></Slider>
+        <div className="speech-actions">
+          <span className="tags">
+            <span className="item">{props.format}</span>
+          </span>
+          <span className="duration">
+            {_.round(currentTime, 2) || _.round(duration, 2)}
+          </span>
+          <div className="actions">
+            <Tooltip
+              title={
+                isPlay
+                  ? intl.formatMessage({ id: 'playground.audio.button.stop' })
+                  : intl.formatMessage({ id: 'playground.audio.button.play' })
+              }
+            >
+              <Button
+                disabled={!props.audioUrl || duration === 0}
+                onClick={handlePlay}
+                icon={
+                  isPlay ? (
+                    <PauseCircleOutlined className="font-size-16" />
+                  ) : (
+                    <PlayCircleOutlined className="font-size-16" />
+                  )
+                }
+                type="text"
+                size="small"
+              ></Button>
+            </Tooltip>
+            <Tooltip
+              title={intl.formatMessage({
+                id: 'playground.audio.button.download'
+              })}
+            >
+              <Button
+                disabled={!props.audioUrl || duration === 0}
+                onClick={onDownload}
+                icon={<DownloadOutlined className="font-size-16" />}
+                type="text"
+                size="small"
+              ></Button>
+            </Tooltip>
+          </div>
+        </div>
+      </div>
+    );
+  };
+
   return (
     <div>
       <div className="speech-item">
@@ -150,93 +255,51 @@ const SpeechItem: React.FC<SpeechContentProps> = (props) => {
           style={{ height: 120, width: '100%' }}
           ref={wrapper}
         >
-          <AudioPlayer
-            {...props}
-            audioUrl={props.audioUrl}
-            onReady={handleReady}
-            onFinish={handleOnFinish}
-            onPlay={handleOnPlay}
-            onPause={handleOnPause}
-            onAnalyse={handleOnAnalyse}
-            onAudioprocess={handleOnAudioprocess}
-            ref={ref}
-          ></AudioPlayer>
-          {/* <RawAudioPlayer
-            {...props}
-            url={props.audioUrl}
-            onReady={handleReady}
-            onEnded={handleOnFinish}
-            onPlay={handleOnPlay}
-            onPause={handleOnPause}
-            onAnalyse={handleOnAnalyse}
-            onAudioProcess={handleOnAudioprocess}
-            ref={ref}
-          ></RawAudioPlayer> */}
-          {isPlay && (
-            <AudioAnimation
-              maxBarCount={100}
-              amplitude={60}
-              fixedHeight={true}
-              height={120}
-              width={800}
-              analyserData={audioChunks}
-            ></AudioAnimation>
-          )}
-        </div>
-      </div>
-      <Slider
-        className="slider-progress"
-        value={currentTime}
-        max={duration}
-        step={0.01}
-        onChange={handleSliderChange}
-      ></Slider>
-      <div className="speech-actions">
-        <span className="tags">
-          <span className="item">{props.format}</span>
-        </span>
-        <span className="duration">
-          {_.round(currentTime, 2) || _.round(duration, 2)}
-        </span>
-        <div className="actions">
-          <Tooltip
-            title={
-              isPlay
-                ? intl.formatMessage({ id: 'playground.audio.button.stop' })
-                : intl.formatMessage({ id: 'playground.audio.button.play' })
-            }
-          >
-            <Button
-              disabled={!props.audioUrl || duration === 0}
-              onClick={handlePlay}
-              icon={
-                isPlay ? (
-                  <PauseCircleOutlined className="font-size-16" />
-                ) : (
-                  <PlayCircleOutlined className="font-size-16" />
-                )
-              }
-              type="text"
-              size="small"
-            ></Button>
-          </Tooltip>
-          <Tooltip
-            title={intl.formatMessage({
-              id: 'playground.audio.button.download'
-            })}
-          >
-            <Button
-              disabled={!props.audioUrl || duration === 0}
-              onClick={onDownload}
-              icon={<DownloadOutlined className="font-size-16" />}
-              type="text"
-              size="small"
-            ></Button>
-          </Tooltip>
+          <>
+            {/* {!isPCMStream && (
+              <AudioPlayer
+                {...props}
+                audioUrl={props.audioUrl}
+                onReady={handleReady}
+                onFinish={handleOnFinish}
+                onPlay={handleOnPlay}
+                onPause={handleOnPause}
+                onAnalyse={handleOnAnalyse}
+                onAudioprocess={handleOnAudioprocess}
+                ref={ref}
+              ></AudioPlayer>
+            )} */}
+            {!isPCMStream && (
+              <RawAudioPlayer
+                {...props}
+                url={props.audioUrl}
+                onReady={handleReady}
+                onEnded={handleOnFinish}
+                onPlay={handleOnPlay}
+                onPause={handleOnPause}
+                onAnalyse={handleOnAnalyse}
+                onAudioProcess={handleOnAudioprocess}
+                ref={ref}
+              ></RawAudioPlayer>
+            )}
+            {isPlay &&
+              (props.analyserData?.analyser?.current ||
+                audioChunks.analyser?.current) && (
+                <AudioAnimation
+                  maxBarCount={100}
+                  amplitude={60}
+                  fixedHeight={true}
+                  height={120}
+                  width={800}
+                  analyserData={props.analyserData || audioChunks}
+                ></AudioAnimation>
+              )}
+          </>
         </div>
       </div>
+      {renderPlayerActions()}
     </div>
   );
 };
 
-export default React.memo(SpeechItem);
+export default SpeechItem;

src/pages/benchmark/index.tsx

@@ -9,6 +9,7 @@ import { useQueryModelList } from '@/pages/llmodels/services/use-query-model-lis
 import { useIntl, useNavigate } from '@umijs/max';
 import { useMemoizedFn } from 'ahooks';
 import { ConfigProvider, Table, message } from 'antd';
+import { createStyles } from 'antd-style';
 import _ from 'lodash';
 import { useEffect } from 'react';
 import NoResult from '../_components/no-result';
@@ -35,6 +36,23 @@ import useQueryDataset from './services/use-query-dataset';
 import useQueryProfiles from './services/use-query-profiles';
 import useStopBenchmark from './services/use-stop-benchmark';
 
+const useStyle = createStyles(({ css, token }) => {
+  const antCls = '.ant';
+  return {
+    customTable: css`
+      ${antCls}-table {
+        ${antCls}-table-container {
+          ${antCls}-table-body,
+          ${antCls}-table-content {
+            scrollbar-width: thin;
+            scrollbar-color: var(--color-scrollbar-thumb) transparent;
+          }
+        }
+      }
+    `
+  };
+});
+
 const Benchmark: React.FC = () => {
   const {
     dataSource,
@@ -58,6 +76,7 @@ const Benchmark: React.FC = () => {
     watch: true,
     contentForDelete: 'menu.models.benchmark'
   });
+  const { styles } = useStyle();
   const intl = useIntl();
   const navigate = useNavigate();
   const { openBenchmarkModal, closeBenchmarkModal, openBenchmarkModalStatus } =
@@ -242,13 +261,14 @@ const Benchmark: React.FC = () => {
           <Table
             tableLayout="fixed"
             columns={columns}
+            className={styles.customTable}
             dataSource={dataSource.dataList}
             rowSelection={rowSelection}
             loading={dataSource.loading}
             sortDirections={TABLE_SORT_DIRECTIONS}
             showSorterTooltip={false}
             rowKey="id"
-            scroll={{ x: 1260 }}
+            scroll={{ x: 1200 }}
             onChange={handleTableChange}
             pagination={{
               showSizeChanger: true,

src/pages/playground/apis/index.ts

@@ -155,16 +155,8 @@ export const textToSpeech = async (params: any, options?: any) => {
   }
 
   const audioBlob = await res.blob();
-  if (audioBlob?.type?.indexOf('audio') === -1) {
-    return {
-      url: '',
-      type: ''
-    };
-  }
-  const audioUrl = audioBlob.size > 0 ? URL.createObjectURL(audioBlob) : '';
   return {
-    url: audioUrl,
-    type: audioBlob.type
+    audioBlob
   };
 };
 

src/pages/playground/components/params-fields.tsx

@@ -55,8 +55,10 @@ const ParamsFields: React.FC<ParamsFieldsProps> = ({ paramsConfig = [] }) => {
               'formItemAttrs',
               'dependencies',
               'disabledConfig',
-              'description'
+              'description',
+              'initAttrs'
             ])}
+            {...item.initAttrs?.(meta)}
           />
         </Form.Item>
       );
@@ -93,7 +95,8 @@ const ParamsFields: React.FC<ParamsFieldsProps> = ({ paramsConfig = [] }) => {
                 'formItemAttrs',
                 'dependencies',
                 'disabledConfig',
-                'description'
+                'description',
+                'initAttrs'
               ])}
               {...item.initAttrs?.(meta)}
             />

src/pages/playground/speech/audio/pcm-player-workerlet.ts

@@ -0,0 +1,94 @@
+const workletCode = `class PCMPlayerProcessor extends AudioWorkletProcessor {
+  constructor() {
+    super();
+
+    this.capacity = 24000 * 2; // 2 seconds buffer
+    this.jitterThreshold = 24000 * 0.1; // 100ms
+
+    this.buffer = new Float32Array(this.capacity);
+
+    this.readIndex = 0;
+    this.writeIndex = 0;
+    this.size = 0;
+
+    this.streamEnded = false;
+    this.completionNotified = false;
+
+    this.port.onmessage = (event) => {
+      const { type, data } = event.data || {};
+
+      if (type === 'push' && data) {
+        this.push(data);
+      }
+
+      if (type === 'clear') {
+        this.readIndex = 0;
+        this.writeIndex = 0;
+        this.size = 0;
+        this.streamEnded = false;
+        this.completionNotified = false;
+      }
+
+      if (type === 'end-stream') {
+        this.streamEnded = true;
+      }
+    };
+  }
+
+  push(data) {
+    for (let i = 0; i < data.length; i++) {
+      if (this.size >= this.capacity) {
+        // backpressure: drop oldest data
+        this.readIndex = (this.readIndex + 1) % this.capacity;
+        this.size--;
+      }
+
+      this.buffer[this.writeIndex] = data[i];
+
+      this.writeIndex = (this.writeIndex + 1) % this.capacity;
+      this.size++;
+    }
+  }
+
+  process(inputs, outputs) {
+    const output = outputs[0];
+    if (!output) return true;
+
+    const channel = output[0];
+
+    for (let i = 0; i < channel.length; i++) {
+      // When stream has ended, ignore jitter threshold and play remaining data
+      if (!this.streamEnded && this.size < this.jitterThreshold) {
+        channel[i] = 0;
+        continue;
+      }
+
+      if (this.size === 0) {
+        channel[i] = 0;
+        continue;
+      }
+
+      channel[i] = this.buffer[this.readIndex];
+
+      this.readIndex = (this.readIndex + 1) % this.capacity;
+      this.size--;
+    }
+
+    // Check if playback is complete
+    if (this.streamEnded && this.size === 0 && !this.completionNotified) {
+      console.log('PCM playback complete, notifying main thread');
+      this.completionNotified = true;
+      this.port.postMessage({ type: 'playback-complete' });
+    }
+
+    return true;
+  }
+}
+
+registerProcessor('pcm-player', PCMPlayerProcessor);`;
+
+export const workerletUrl = (): string => {
+  const blob = new Blob([workletCode], { type: 'application/javascript' });
+  const workletUrl = URL.createObjectURL(blob);
+  return workletUrl;
+};

src/pages/playground/speech/forms/tts-advance.tsx

@@ -128,17 +128,6 @@ const TTSAdvanceConfig: React.FC = () => {
           })}
         ></CheckboxField>
       </Form.Item>
-      <Form.Item
-        name="stream"
-        valuePropName="checked"
-        style={{ marginBottom: 8 }}
-      >
-        <CheckboxField
-          label={intl.formatMessage({
-            id: 'playground.params.streamMode'
-          })}
-        ></CheckboxField>
-      </Form.Item>
     </>
   );
 };

src/pages/playground/speech/forms/tts-form.tsx

@@ -1,4 +1,5 @@
 import AutoComplete from '@/components/seal-form/auto-complete';
+import CheckboxField from '@/components/seal-form/checkbox-field';
 import SealSelect from '@/components/seal-form/seal-select';
 import CollapsePanel from '@/pages/_components/collapse-panel';
 import { getLocale, useIntl, useSearchParams } from '@umijs/max';
@@ -13,7 +14,7 @@ import React, {
 } from 'react';
 import ModelSelect from '../../components/model-select';
 import ParamsFields from '../../components/params-fields';
-import { FormContext } from '../../config/form-context';
+import { FormContext, useFormContext } from '../../config/form-context';
 import { TTSAdvancedParamsConfig } from '../params-config';
 import AdvanceConfig from './tts-advance';
 
@@ -35,6 +36,7 @@ type ParamsSettingsProps = {
 
 const ParamsSettings: React.FC<ParamsSettingsProps> = forwardRef(
   ({ onFinish, onFinishFailed, updatateParams, modelList = [] }, ref) => {
+    const { onValuesChange } = useFormContext();
     const [searchParams] = useSearchParams();
     const modelType = searchParams.get('type') || '';
     const selectModel = searchParams.get('model')
@@ -124,6 +126,14 @@ const ParamsSettings: React.FC<ParamsSettingsProps> = forwardRef(
       if (changeValues.model) {
         return;
       }
+
+      if ('stream' in changeValues && changeValues.stream) {
+        allValues.response_format = 'pcm';
+        form.setFieldsValue({
+          response_format: 'pcm'
+        });
+      }
+
       updatateParams(allValues);
     };
 
@@ -164,15 +174,27 @@ const ParamsSettings: React.FC<ParamsSettingsProps> = forwardRef(
               options={vociceOptions}
             ></AutoComplete>
           </Form.Item>
-          <Form.Item name="response_format">
+          <Form.Item name="response_format" style={{ marginBottom: 8 }}>
             <SealSelect
               label={intl.formatMessage({ id: 'playground.params.format' })}
               options={[
                 { label: 'mp3', value: 'mp3' },
-                { label: 'wav', value: 'wav' }
+                { label: 'wav', value: 'wav' },
+                { label: 'pcm', value: 'pcm' }
               ]}
             ></SealSelect>
           </Form.Item>
+          <Form.Item
+            name="stream"
+            valuePropName="checked"
+            style={{ marginBottom: 8 }}
+          >
+            <CheckboxField
+              label={intl.formatMessage({
+                id: 'playground.params.streamMode'
+              })}
+            ></CheckboxField>
+          </Form.Item>
           <CollapsePanel
             activeKey={activeKey}
             onChange={handleOnCollapse}

src/pages/playground/speech/hooks/use-non-stream-tts.ts

@@ -1,3 +1,4 @@
+import { pcmToWav } from '@/utils/pcm-to-wav';
 import { useCallback, useRef, useState } from 'react';
 import { textToSpeech } from '../../apis';
 import { extractErrorMessage } from '../../config';
@@ -21,6 +22,35 @@ export const useNonStreamTTS = (params?: UseNonStreamTTSParams) => {
   const [error, setError] = useState<any>(null);
   const controllerRef = useRef<AbortController | null>(null);
 
+  const convertPCMToWav = async (ab: Blob) => {
+    let audioBlob = ab;
+    // Convert PCM to WAV for browser playback
+    const arrayBuffer = await audioBlob.arrayBuffer();
+    audioBlob = pcmToWav(arrayBuffer);
+
+    const audioUrl = audioBlob.size > 0 ? URL.createObjectURL(audioBlob) : '';
+    return {
+      url: audioUrl,
+      type: audioBlob.type
+    };
+  };
+
+  // handle non-pcm formats
+  const generateAudioUrl = async (audioBlob: Blob) => {
+    const audioUrl = audioBlob.size > 0 ? URL.createObjectURL(audioBlob) : '';
+    return {
+      url: audioUrl,
+      type: audioBlob.type
+    };
+  };
+
+  const isPCMFormat = (audioBlob: Blob) => {
+    return (
+      audioBlob.type === 'audio/pcm' ||
+      audioBlob.type === 'application/octet-stream'
+    );
+  };
+
   const generate = useCallback(
     async (ttsParams: TTSParams) => {
       try {
@@ -47,7 +77,26 @@ export const useNonStreamTTS = (params?: UseNonStreamTTSParams) => {
           return null;
         }
 
-        params?.onSuccess?.(res);
+        let result = {
+          url: '',
+          type: ''
+        };
+
+        if (res.audioBlob?.type?.indexOf('audio') === -1) {
+          result = {
+            url: '',
+            type: ''
+          };
+        } else if (
+          ttsParams.response_format === 'pcm' ||
+          isPCMFormat(res.audioBlob)
+        ) {
+          result = await convertPCMToWav(res.audioBlob);
+        } else {
+          result = await generateAudioUrl(res.audioBlob);
+        }
+
+        params?.onSuccess?.(result);
         return res;
       } catch (err: any) {
         const res = err?.response?.data;

src/pages/playground/speech/hooks/use-pcm-stream-player.ts

@@ -0,0 +1,172 @@
+import { useMemoizedFn } from 'ahooks';
+import { useRef, useState } from 'react';
+import { workerletUrl } from '../audio/pcm-player-workerlet';
+
+interface Params {
+  sampleRate?: number;
+  numChannels?: number;
+  bitsPerSample?: number;
+  onReady?: () => void;
+  onError?: (err: any) => void;
+  onPlaybackComplete?: () => void;
+}
+
+export const usePCMStreamPlayer = (params?: Params) => {
+  const {
+    sampleRate = 24000,
+    numChannels = 1,
+    bitsPerSample = 16,
+    onReady,
+    onError,
+    onPlaybackComplete
+  } = params || {};
+
+  const [isPlaying, setIsPlaying] = useState(false);
+
+  const audioContextRef = useRef<AudioContext | null>(null);
+  const workletNodeRef = useRef<AudioWorkletNode | null>(null);
+  const analyserRef = useRef<AnalyserNode | null>(null);
+
+  const leftoverRef = useRef<Uint8Array | null>(null);
+  const streamEndedRef = useRef(false);
+
+  const [audioChunks, setAudioChunks] = useState<any>({
+    data: new Uint8Array(128),
+    analyser: null
+  });
+
+  const initialize = useMemoizedFn(async () => {
+    if (audioContextRef.current) return;
+
+    try {
+      const ctx = new AudioContext({ sampleRate });
+
+      const workletUrl = workerletUrl();
+
+      await ctx.audioWorklet.addModule(workletUrl);
+
+      URL.revokeObjectURL(workletUrl);
+      const node = new AudioWorkletNode(ctx, 'pcm-player');
+
+      // Listen for messages from the worklet
+      node.port.onmessage = (event) => {
+        if (event.data.type === 'playback-complete') {
+          setIsPlaying(false);
+          onPlaybackComplete?.();
+        }
+      };
+
+      const analyser = ctx.createAnalyser();
+      analyser.fftSize = 512;
+
+      node.connect(analyser);
+      analyser.connect(ctx.destination);
+
+      audioContextRef.current = ctx;
+      workletNodeRef.current = node;
+      analyserRef.current = analyser;
+
+      setAudioChunks({
+        data: new Uint8Array(analyser.frequencyBinCount),
+        analyser: analyserRef
+      });
+
+      streamEndedRef.current = false;
+
+      onReady?.();
+    } catch (err) {
+      onError?.(err);
+    }
+  });
+
+  const convertPCM = (pcmData: Uint8Array) => {
+    let pcm = pcmData;
+    const bytesPerSample = bitsPerSample / 8;
+    const frameSize = bytesPerSample * numChannels;
+
+    // If there's leftover data from the previous chunk, prepend it to the current chunk
+    if (leftoverRef.current) {
+      const merged = new Uint8Array(leftoverRef.current.length + pcm.length);
+      merged.set(leftoverRef.current);
+      merged.set(pcm, leftoverRef.current.length);
+      pcm = merged;
+    }
+
+    const usableLength = Math.floor(pcm.length / frameSize) * frameSize;
+
+    leftoverRef.current = pcm.slice(usableLength);
+
+    const usable = pcm.slice(0, usableLength);
+
+    const view = new DataView(
+      usable.buffer,
+      usable.byteOffset,
+      usable.byteLength
+    );
+
+    const numSamples = usable.length / 2;
+
+    const float32 = new Float32Array(numSamples);
+
+    for (let i = 0; i < numSamples; i++) {
+      float32[i] = view.getInt16(i * 2, true) / 32768;
+    }
+
+    return float32;
+  };
+
+  const addChunk = useMemoizedFn((pcmChunk: Uint8Array) => {
+    const node = workletNodeRef.current;
+    if (!node) return;
+
+    setIsPlaying(true);
+    const floatData = convertPCM(pcmChunk);
+
+    if (!floatData.length) return;
+
+    node.port.postMessage({
+      type: 'push',
+      data: floatData
+    });
+  });
+
+  const stop = useMemoizedFn(() => {
+    const ctx = audioContextRef.current;
+    if (!ctx) return;
+
+    workletNodeRef.current?.port.postMessage({ type: 'clear' });
+
+    ctx.close();
+
+    audioContextRef.current = null;
+    workletNodeRef.current = null;
+    analyserRef.current = null;
+    streamEndedRef.current = false;
+
+    setIsPlaying(false);
+  });
+
+  const cleanup = useMemoizedFn(() => {
+    stop();
+  });
+
+  const endStream = useMemoizedFn(() => {
+    const node = workletNodeRef.current;
+    if (!node) return;
+
+    streamEndedRef.current = true;
+    node.port.postMessage({ type: 'end-stream' });
+  });
+
+  return {
+    initialize,
+    addChunk,
+    stop,
+    cleanup,
+    endStream,
+    isPlaying,
+    setIsPlaying,
+    setAudioChunks,
+    audioChunks
+  };
+};

src/pages/playground/speech/hooks/use-stream-tts.ts

@@ -1,13 +1,18 @@
 import { fetchChunkedData } from '@/utils/fetch-chunk-data';
+import { pcmToWav } from '@/utils/pcm-to-wav';
+import { useMemoizedFn } from 'ahooks';
 import { useCallback, useRef, useState } from 'react';
 import { AUDIO_TEXT_TO_SPEECH_API } from '../../apis';
 import { extractErrorMessage } from '../../config';
+import { usePCMStreamPlayer } from './use-pcm-stream-player';
 
 interface UseStreamTTSParams {
   onChunk?: (chunk: ArrayBuffer) => void;
   onComplete?: (audioUrl: string) => void; // Return complete audio URL when done
   onError?: (error: any) => void;
+  onUrlReady?: (url: string) => void; // Called when the stream URL is ready
   autoPlay?: boolean;
+  playerRef?: React.RefObject<any>; // Reference to the player for controlling playback
 }
 
 interface TTSParams {
@@ -20,262 +25,325 @@ interface TTSParams {
   [key: string]: any;
 }
 
-/**
- * Audio chunk queue manager for smooth playback
- * Uses a single Audio element for better performance and seamless playback
- */
-class AudioQueue {
-  private queue: Blob[] = [];
-  private audioElement: HTMLAudioElement;
-  private isPlaying = false;
-  private currentIndex = 0;
-  private onComplete?: () => void;
-  private streamEnded = false;
-  private minBufferSize = 2; // Minimum chunks to buffer before starting playback
-  private maxQueueSize = 10; // Maximum queue size to prevent memory issues
-  private currentBlobUrl: string | null = null;
+// MediaSource codec mapping for different formats
+const MEDIA_SOURCE_CODECS: Record<string, string> = {
+  mp4: 'audio/mp4; codecs="mp4a.40.2"',
+  webm: 'audio/webm; codecs="opus"',
+  ogg: 'audio/ogg; codecs="opus"',
+  opus: 'audio/webm; codecs="opus"',
+  pcm: 'audio/pcm; codecs=pcm'
+};
 
-  constructor(onComplete?: () => void) {
-    this.onComplete = onComplete;
-    // Create a single Audio element for the entire playback session
-    this.audioElement = new Audio();
-    this.setupAudioListeners();
-  }
-
-  private setupAudioListeners() {
-    this.audioElement.onended = () => {
-      this.cleanupCurrentBlob();
-      this.playNext();
-    };
-
-    this.audioElement.onerror = () => {
-      console.error('Audio playback error');
-      this.cleanupCurrentBlob();
-      this.playNext();
-    };
-  }
-
-  private cleanupCurrentBlob() {
-    if (this.currentBlobUrl) {
-      URL.revokeObjectURL(this.currentBlobUrl);
-      this.currentBlobUrl = null;
-    }
-  }
-
-  addChunk(chunk: Blob) {
-    this.queue.push(chunk);
-
-    // Start playback if we have enough buffer
-    if (!this.isPlaying && this.queue.length >= this.minBufferSize) {
-      this.playNext();
-    }
-  }
-
-  // Check if queue is full (for backpressure)
-  isFull(): boolean {
-    return this.queue.length - this.currentIndex >= this.maxQueueSize;
-  }
-
-  // Get current queue size (unplayed chunks)
-  getQueueSize(): number {
-    return this.queue.length - this.currentIndex;
-  }
-
-  private async playNext() {
-    if (this.currentIndex >= this.queue.length) {
-      // If stream has ended and no more chunks, complete
-      if (this.streamEnded) {
-        this.isPlaying = false;
-        this.onComplete?.();
-      }
-      return;
-    }
-
-    this.isPlaying = true;
-    const chunk = this.queue[this.currentIndex];
-    this.currentIndex++;
-
-    // Reuse the same Audio element, just update the src
-    this.cleanupCurrentBlob();
-    this.currentBlobUrl = URL.createObjectURL(chunk);
-    this.audioElement.src = this.currentBlobUrl;
-
-    try {
-      await this.audioElement.play();
-    } catch (error) {
-      console.error('Failed to play audio chunk:', error);
-      this.cleanupCurrentBlob();
-      this.playNext();
-    }
-  }
-
-  markStreamEnded() {
-    this.streamEnded = true;
-    // If not playing and has remaining chunks, start playing
-    if (!this.isPlaying && this.currentIndex < this.queue.length) {
-      this.playNext();
-    }
-  }
-
-  stop() {
-    this.isPlaying = false;
-    this.audioElement.pause();
-    this.cleanupCurrentBlob();
-  }
-
-  clear() {
-    this.stop();
-    this.queue = [];
-    this.currentIndex = 0;
-    this.streamEnded = false;
-  }
-}
+// Check if format supports MediaSource API
+const supportsMediaSource = (format: string): boolean => {
+  if (!window.MediaSource) return false;
+  const codec = MEDIA_SOURCE_CODECS[format];
+  return codec ? MediaSource.isTypeSupported(codec) : false;
+};
 
 export const useStreamTTS = (params?: UseStreamTTSParams) => {
   const [loading, setLoading] = useState(false);
   const [error, setError] = useState<any>(null);
   const [progress, setProgress] = useState(0);
+  const [streamUrl, setStreamUrl] = useState<string>('');
   const controllerRef = useRef<AbortController | null>(null);
-  const audioQueueRef = useRef<AudioQueue | null>(null);
+  const mediaSourceRef = useRef<MediaSource | null>(null);
+  const sourceBufferRef = useRef<SourceBuffer | null>(null);
+  const queueRef = useRef<Uint8Array[]>([]);
+  const isAppendingRef = useRef(false);
+  const allChunksRef = useRef<Uint8Array[]>([]);
+  const [isPCM, setIsPCM] = useState(false);
+  const completeAudioUrlRef = useRef<string>('');
 
-  const generate = useCallback(
-    async (ttsParams: TTSParams) => {
-      try {
-        setLoading(true);
-        setError(null);
-        setProgress(0);
+  // PCM stream player instance
+  const pcmPlayer = usePCMStreamPlayer({
+    onError: (error) => {
+      console.error('PCM player error:', error);
+      params?.onError?.(error);
+    },
+    onPlaybackComplete: () => {
+      console.log('PCM playback complete');
+      // Create complete audio URL from all chunks when playback finishes
+      if (allChunksRef.current.length > 0) {
+        const totalLength = allChunksRef.current.reduce(
+          (acc, chunk) => acc + chunk.length,
+          0
+        );
+        const pcmData = new Uint8Array(totalLength);
+        let offset = 0;
+        for (const chunk of allChunksRef.current) {
+          pcmData.set(chunk, offset);
+          offset += chunk.length;
+        }
 
-        // Abort previous request if exists
-        controllerRef.current?.abort();
-        audioQueueRef.current?.clear();
+        // Convert combined PCM data to WAV and create URL
+        const wavBlob = pcmToWav(pcmData.buffer);
+        const completeUrl = URL.createObjectURL(wavBlob);
+        params?.onComplete?.(completeUrl);
+        // when the stream ends, should clear the audio analyer data, because the the audio elment will play the complete audio, and the analyser will generate new.
+        pcmPlayer.setAudioChunks(null);
+      }
+    }
+  });
 
-        controllerRef.current = new AbortController();
-        const signal = controllerRef.current.signal;
+  const processQueue = useCallback(() => {
+    if (
+      isAppendingRef.current ||
+      queueRef.current.length === 0 ||
+      !sourceBufferRef.current ||
+      sourceBufferRef.current.updating
+    ) {
+      return;
+    }
 
-        // Create audio queue for smooth playback
-        audioQueueRef.current = new AudioQueue();
+    isAppendingRef.current = true;
+    const chunk = queueRef.current.shift()!;
+    try {
+      sourceBufferRef.current.appendBuffer(chunk.buffer as ArrayBuffer);
+    } catch (error) {
+      console.error('Failed to append buffer:', error);
+      isAppendingRef.current = false;
+    }
+  }, []);
 
-        // Add stream parameter
-        const streamParams = {
-          ...ttsParams,
-          stream: true
-        };
+  const generate = useMemoizedFn(async (ttsParams: TTSParams) => {
+    try {
+      setLoading(true);
+      setError(null);
+      setProgress(0);
+      setIsPCM(ttsParams.response_format === 'pcm');
+      allChunksRef.current = [];
 
-        const result = await fetchChunkedData({
-          url: AUDIO_TEXT_TO_SPEECH_API,
-          data: streamParams,
-          signal
-        });
+      // Abort previous request if exists
+      controllerRef.current?.abort();
 
-        if ('error' in result) {
-          const errorMessage = extractErrorMessage(result.data);
-          setError({
-            error: true,
-            errorMessage
+      // Clean up previous PCM player
+      pcmPlayer.stop();
+
+      // Clean up previous MediaSource/URL
+      if (streamUrl) {
+        URL.revokeObjectURL(streamUrl);
+        setStreamUrl('');
+      }
+
+      if (mediaSourceRef.current) {
+        if (mediaSourceRef.current.readyState === 'open') {
+          mediaSourceRef.current.endOfStream();
+        }
+        mediaSourceRef.current = null;
+      }
+
+      sourceBufferRef.current = null;
+      queueRef.current = [];
+      isAppendingRef.current = false;
+
+      controllerRef.current = new AbortController();
+      const signal = controllerRef.current.signal;
+
+      const format = ttsParams.response_format || 'mp3';
+      const isPCM = format === 'pcm';
+      const useMediaSource = !isPCM && supportsMediaSource(format);
+
+      let url = '';
+      console.log(
+        'format:',
+        format,
+        'isPCM:',
+        isPCM,
+        'useMediaSource:',
+        useMediaSource
+      );
+
+      if (isPCM) {
+        // PCM format: use Web Audio API for playback
+        pcmPlayer.initialize();
+        // Create a virtual URL identifier for PCM stream (not an actual URL since we're using Web Audio API)
+        url = 'pcm-stream://playing';
+        setStreamUrl(url);
+        params?.onUrlReady?.(url);
+      } else if (useMediaSource) {
+        // Use MediaSource API for supported formats
+        const mediaSource = new MediaSource();
+        mediaSourceRef.current = mediaSource;
+        url = URL.createObjectURL(mediaSource);
+        console.log('Created MediaSource URL:', url);
+        setStreamUrl(url);
+        params?.onUrlReady?.(url); // Notify that URL is ready
+        console.log('Called onUrlReady with URL:', url);
+
+        // Wait for MediaSource to be ready
+        await new Promise<void>((resolve, reject) => {
+          const handleSourceOpen = () => {
+            try {
+              const mimeType = MEDIA_SOURCE_CODECS[format];
+              const sourceBuffer = mediaSource.addSourceBuffer(mimeType);
+              sourceBufferRef.current = sourceBuffer;
+
+              sourceBuffer.addEventListener('updateend', () => {
+                isAppendingRef.current = false;
+                processQueue();
+              });
+
+              sourceBuffer.addEventListener('error', (e) => {
+                console.error('SourceBuffer error:', e);
+              });
+
+              resolve();
+            } catch (error) {
+              reject(error);
+            }
+          };
+
+          mediaSource.addEventListener('sourceopen', handleSourceOpen, {
+            once: true
           });
-          params?.onError?.(errorMessage);
-          return;
-        }
 
-        const { reader } = result;
-
-        if (!reader) {
-          throw new Error('Failed to get reader from response');
-        }
-
-        // Read stream data
-        let audioChunks: Uint8Array[] = [];
-        let allAudioChunks: Uint8Array[] = []; // Collect all chunks for final URL
-        let chunkCount = 0;
-
-        while (true) {
-          // Backpressure: wait if queue is full
-          while (audioQueueRef.current?.isFull()) {
-            await new Promise((resolve) => {
-              setTimeout(resolve, 100);
-            });
-          }
-
-          const { done, value } = await reader.read();
-
-          if (done) {
-            // Process remaining chunks for playback
-            if (audioChunks.length > 0) {
-              const blob = new Blob(audioChunks as any, {
-                type: `audio/${ttsParams.response_format || 'mp3'}`
-              });
-              audioQueueRef.current?.addChunk(blob);
-              params?.onChunk?.(blob.arrayBuffer() as any);
+          // Timeout fallback
+          setTimeout(() => {
+            if (mediaSource.readyState !== 'open') {
+              reject(new Error('MediaSource failed to open'));
             }
-            audioQueueRef.current?.markStreamEnded();
+          }, 5000);
+        });
+      }
 
-            // Create complete audio URL from all chunks
-            if (allAudioChunks.length > 0) {
-              const completeBlob = new Blob(allAudioChunks as any, {
-                type: `audio/${ttsParams.response_format || 'mp3'}`
-              });
-              const completeUrl = URL.createObjectURL(completeBlob);
-              params?.onComplete?.(completeUrl);
-            }
-            break;
-          }
+      // Add stream parameter
+      const streamParams = {
+        ...ttsParams,
+        stream: true
+      };
 
-          if (value) {
-            audioChunks.push(value);
-            allAudioChunks.push(value); // Keep all chunks for final URL
-            chunkCount++;
+      const result = await fetchChunkedData({
+        url: AUDIO_TEXT_TO_SPEECH_API,
+        data: streamParams,
+        signal
+      });
 
-            // For fast API responses, batch chunks to avoid too many small audio elements
-            // Create a blob every N chunks or when chunk size exceeds threshold
-            const totalSize = audioChunks.reduce(
-              (sum, chunk) => sum + chunk.length,
-              0
-            );
-            const shouldFlush = chunkCount >= 5 || totalSize >= 50000; // ~50KB threshold
-
-            if (shouldFlush) {
-              const blob = new Blob(audioChunks as any, {
-                type: `audio/${ttsParams.response_format || 'mp3'}`
-              });
-              audioQueueRef.current?.addChunk(blob);
-              params?.onChunk?.(blob.arrayBuffer() as any);
-
-              audioChunks = [];
-              chunkCount = 0;
-              setProgress((prev) => prev + 1);
-            }
-          }
-        }
-      } catch (err: any) {
-        if (err.name === 'AbortError') {
-          console.log('Stream aborted');
-          return;
-        }
-
-        const errorMessage = err?.message || 'Stream processing failed';
+      if ('error' in result) {
+        const errorMessage = extractErrorMessage(result.data);
         setError({
           error: true,
           errorMessage
         });
         params?.onError?.(errorMessage);
-      } finally {
-        setLoading(false);
+        return;
       }
-    },
-    [params]
-  );
+
+      const { reader } = result;
+
+      if (!reader) {
+        throw new Error('Failed to get reader from response');
+      }
+
+      // Read stream data
+      while (true) {
+        const { done, value } = await reader.read();
+
+        if (done) {
+          if (useMediaSource) {
+            // Wait for all chunks to be appended before ending stream
+            await new Promise<void>((resolve) => {
+              const checkQueue = () => {
+                if (queueRef.current.length === 0 && !isAppendingRef.current) {
+                  if (mediaSourceRef.current?.readyState === 'open') {
+                    try {
+                      mediaSourceRef.current.endOfStream();
+                      setLoading(false);
+                    } catch (error) {
+                      console.error('Failed to end stream:', error);
+                    }
+                  }
+                  resolve();
+                } else {
+                  setTimeout(checkQueue, 100);
+                }
+              };
+              checkQueue();
+            });
+          }
+
+          // Handle completion based on format
+          if (allChunksRef.current.length > 0) {
+            if (isPCM) {
+              // PCM format: notify the player that streaming has ended
+              // onComplete will be called by the player when playback finishes
+              pcmPlayer.endStream();
+            } else {
+              const completeBlob = new Blob(allChunksRef.current as any, {
+                type: `audio/${format}`
+              });
+              const completeUrl = URL.createObjectURL(completeBlob);
+            }
+          }
+          break;
+        }
+
+        if (value) {
+          allChunksRef.current.push(value);
+          if (isPCM) {
+            // PCM chunks are sent directly to the player for real-time playback
+            pcmPlayer.addChunk(value);
+          } else if (useMediaSource) {
+            queueRef.current.push(value);
+            processQueue();
+          }
+
+          params?.onChunk?.(value.buffer);
+          setProgress((prev) => prev + 1);
+        }
+      }
+    } catch (err: any) {
+      if (err.name === 'AbortError') {
+        console.log('Stream aborted');
+        return;
+      }
+
+      const errorMessage = err?.message || 'Stream processing failed';
+      setError({
+        error: true,
+        errorMessage
+      });
+      params?.onError?.(errorMessage);
+    } finally {
+      setLoading(false);
+    }
+  });
 
   const abort = useCallback(() => {
     controllerRef.current?.abort();
-    audioQueueRef.current?.stop();
+    pcmPlayer.stop();
+    if (streamUrl) {
+      URL.revokeObjectURL(streamUrl);
+    }
+    if (mediaSourceRef.current?.readyState === 'open') {
+      try {
+        mediaSourceRef.current.endOfStream();
+      } catch (error) {
+        console.error('Failed to end stream on abort:', error);
+      }
+    }
     setLoading(false);
-  }, []);
+  }, [streamUrl, pcmPlayer]);
+
+  const play = useCallback(() => {
+    params?.playerRef?.current?.play();
+  }, [params?.playerRef]);
+
+  const pause = useCallback(() => {
+    params?.playerRef?.current?.pause();
+  }, [params?.playerRef]);
 
   return {
     generate,
     abort,
     loading,
     error,
-    progress
+    progress,
+    streamUrl,
+    audioChunks: isPCM ? pcmPlayer.audioChunks : undefined,
+    isPlaying: pcmPlayer.isPlaying,
+    pcmPlayer,
+    play,
+    pause
   };
 };

src/pages/playground/speech/params-config.ts

@@ -80,7 +80,6 @@ export const TTSAdvancedParamsConfig: ParamsSchema[] = [
       isLocalized: true
     },
     attrs: {
-      allowClear: true,
       step: 1,
       min: 0,
       max: 4096

src/pages/playground/speech/tts.tsx

@@ -58,8 +58,18 @@ const GroundTTS: React.FC<MessageProps> = forwardRef((props, ref) => {
   const checkvalueRef = useRef<any>(true);
   const [currentPrompt, setCurrentPrompt] = useState<string>('');
   const formRef = useRef<any>(null);
+  const [playingStream, setPlayingStream] = useState(false);
+
+  const handleOnPlay = () => {
+    // TODO
+    console.log('Play audio');
+  };
+
+  const handleOnPause = () => {
+    // TODO
+    console.log('Pause audio');
+  };
 
-  // Initialize non-stream TTS hook
   const nonStreamTTS = useNonStreamTTS({
     onSuccess: (result) => {
       setMessageList([
@@ -83,11 +93,25 @@ const GroundTTS: React.FC<MessageProps> = forwardRef((props, ref) => {
     }
   });
 
-  // Initialize stream TTS hook
+  const playerRef = useRef<any>(null);
+
   const streamTTS = useStreamTTS({
     autoPlay: checkvalueRef.current,
+    playerRef: playerRef,
+    onUrlReady: (url) => {
+      console.log('onUrlReady called with URL:', url);
+      // Update messageList with stream URL when it's ready
+      setMessageList((prev) => {
+        if (prev.length > 0) {
+          const updated = [...prev];
+          updated[updated.length - 1].audioUrl = url;
+          return updated;
+        }
+        return prev;
+      });
+    },
     onComplete: (audioUrl) => {
-      // Update messageList with complete audio URL when stream finishes
+      console.log('onComplete called with audioUrl:', audioUrl);
       setMessageList((prev) => {
         if (prev.length > 0) {
           const updated = [...prev];
@@ -96,7 +120,7 @@ const GroundTTS: React.FC<MessageProps> = forwardRef((props, ref) => {
         }
         return prev;
       });
-      console.log('Stream playback completed');
+      setPlayingStream(false);
     },
     onError: (error) => {
       setTokenResult({
@@ -104,9 +128,12 @@ const GroundTTS: React.FC<MessageProps> = forwardRef((props, ref) => {
         errorMessage: error
       });
       setMessageList([]);
+      setPlayingStream(false);
     }
   });
 
+  console.log('isPlaying:', streamTTS.isPlaying);
+
   useImperativeHandle(ref, () => {
     return {
       viewCode() {
@@ -180,10 +207,12 @@ const GroundTTS: React.FC<MessageProps> = forwardRef((props, ref) => {
       };
 
       setParams(params);
+      console.log('submitMessage params:', streamTTS.isPlaying);
 
       // Choose stream or non-stream based on parameters
       if (parameters.stream) {
-        // Stream mode: audio will play in real-time
+        setPlayingStream(true);
+        // Stream mode: create initial message entry, URL will be set via onUrlReady callback
         setMessageList([
           {
             input: inputText,
@@ -191,8 +220,8 @@ const GroundTTS: React.FC<MessageProps> = forwardRef((props, ref) => {
             format: parameters.response_format,
             speed: parameters.speed,
             uid: messageId.current,
-            autoplay: checkvalueRef.current,
-            audioUrl: '' // No URL for stream mode, audio plays in real-time
+            autoplay: false,
+            audioUrl: '' // Will be updated via onUrlReady callback
           }
         ]);
         await streamTTS.generate(params);
@@ -202,6 +231,7 @@ const GroundTTS: React.FC<MessageProps> = forwardRef((props, ref) => {
       }
     } catch (error: any) {
       console.log('error:', error);
+      setPlayingStream(false);
       setTokenResult({
         error: true,
         errorMessage: error?.message || 'Unknown error'
@@ -252,7 +282,16 @@ const GroundTTS: React.FC<MessageProps> = forwardRef((props, ref) => {
           >
             <div className="content" style={{ maxWidth: 1000 }}>
               {messageList.length ? (
-                <SpeechContent dataList={messageList} loading={loading} />
+                <SpeechContent
+                  dataList={messageList}
+                  loading={loading}
+                  onPlay={handleOnPlay}
+                  onPause={handleOnPause}
+                  playerRef={playerRef}
+                  isPlaying={playingStream}
+                  isStream={parameters.stream}
+                  analyserData={streamTTS.audioChunks}
+                />
               ) : (
                 <div className="flex-column font-size-14 flex-center gap-20">
                   <span>

src/utils/pcm-to-wav.ts

@@ -0,0 +1,54 @@
+function writeString(view: DataView, offset: number, string: string) {
+  for (let i = 0; i < string.length; i++) {
+    view.setUint8(offset + i, string.charCodeAt(i));
+  }
+}
+
+/**
+ * Convert PCM audio data to WAV format
+ * @param pcmData - Raw PCM audio data as ArrayBuffer
+ * @param sampleRate - Sample rate (default: 24000 Hz, common for TTS)
+ * @param numChannels - Number of channels (default: 1 for mono)
+ * @param bitsPerSample - Bits per sample (default: 16)
+ * @returns WAV format audio as Blob
+ */
+export function pcmToWav(
+  pcmData: ArrayBuffer,
+  sampleRate: number = 24000,
+  numChannels: number = 1,
+  bitsPerSample: number = 16
+): Blob {
+  const pcmBytes = new Uint8Array(pcmData);
+  const dataLength = pcmBytes.length;
+
+  // WAV file header size
+  const headerSize = 44;
+  const wavBuffer = new ArrayBuffer(headerSize + dataLength);
+  const view = new DataView(wavBuffer);
+
+  // Write WAV header
+  // "RIFF" chunk descriptor
+  writeString(view, 0, 'RIFF');
+  view.setUint32(4, 36 + dataLength, true); // File size - 8
+  writeString(view, 8, 'WAVE');
+
+  // "fmt " sub-chunk
+  writeString(view, 12, 'fmt ');
+  view.setUint32(16, 16, true); // Subchunk1Size (16 for PCM)
+  view.setUint16(20, 1, true); // AudioFormat (1 for PCM)
+  view.setUint16(22, numChannels, true); // NumChannels
+  view.setUint32(24, sampleRate, true); // SampleRate
+  view.setUint32(28, sampleRate * numChannels * (bitsPerSample / 8), true); // ByteRate
+  view.setUint16(32, numChannels * (bitsPerSample / 8), true); // BlockAlign
+  view.setUint16(34, bitsPerSample, true); // BitsPerSample
+
+  // "data" sub-chunk
+  writeString(view, 36, 'data');
+  view.setUint32(40, dataLength, true); // Subchunk2Size
+
+  // Write PCM data
+  const wavBytes = new Uint8Array(wavBuffer);
+  wavBytes.set(pcmBytes, headerSize);
+
+  return new Blob([wavBuffer], { type: 'audio/wav' });
+}