diff --git a/examples/Example.h b/examples/Example.h
index 02f22075..bad0177f 100644
--- a/examples/Example.h
+++ b/examples/Example.h
@@ -425,14 +425,14 @@ struct WgWindow : Window
 
     virtual ~WgWindow()
     {
-        wgpuSurfaceRelease(surface);
+        //wgpuSurfaceRelease(surface);
         wgpuInstanceRelease(instance);
     }
 
     void resize() override
     {
         //Set the canvas target and draw on it.
-        verify(static_cast<tvg::WgCanvas*>(canvas)->target(instance, surface, width, height));
+        verify(static_cast<tvg::WgCanvas*>(canvas)->target(nullptr, instance, surface, width, height));
     }
 
     void refresh() override 
diff --git a/inc/thorvg.h b/inc/thorvg.h
index 1a12b4c5..da455613 100644
--- a/inc/thorvg.h
+++ b/inc/thorvg.h
@@ -1774,11 +1774,12 @@ public:
     /**
      * @brief Sets the drawing target for the rasterization.
      *
-     * @param[in] instance WGPUInstance, context for all other wgpu objects.
-     * @param[in] surface WGPUSurface, handle to a presentable surface.
-     * @param[in] w The width of the surface.
-     * @param[in] h The height of the surface.
      * @param[in] device WGPUDevice, a desired handle for the wgpu device. If it is @c nullptr, ThorVG will assign an appropriate device internally.
+     * @param[in] instance WGPUInstance, context for all other wgpu objects.
+     * @param[in] target Either WGPUSurface or WGPUTexture, serving as handles to a presentable surface or texture
+     * @param[in] w The width of the target.
+     * @param[in] h The height of the target.
+     * @param[in] type 0: surface, 1: texture are used as pesentable target
      *
      * @retval Result::InsufficientCondition if the canvas is performing rendering. Please ensure the canvas is synced.
      * @retval Result::NonSupport In case the wg engine is not supported.
@@ -1788,7 +1789,7 @@ public:
      * @see Canvas::viewport()
      * @see Canvas::sync()
      */
-    Result target(void* instance, void* surface, uint32_t w, uint32_t h, void* device = nullptr) noexcept;
+    Result target(void* device, void* instance, void* target, uint32_t w, uint32_t h, int type = 0) noexcept;
 
     /**
      * @brief Creates a new WgCanvas object.
@@ -2148,4 +2149,4 @@ public:
 
 } //namespace
 
-#endif //_THORVG_H_
+#endif //_THORVG_H_
\ No newline at end of file
diff --git a/src/bindings/capi/thorvg_capi.h b/src/bindings/capi/thorvg_capi.h
index 0d3b31a0..ce291511 100644
--- a/src/bindings/capi/thorvg_capi.h
+++ b/src/bindings/capi/thorvg_capi.h
@@ -560,11 +560,12 @@ TVG_API Tvg_Canvas* tvg_wgcanvas_create(void);
 /*!
 * \brief Sets the drawing target for the rasterization.
 *
-* \param[in] instance WGPUInstance, context for all other wgpu objects.
-* \param[in] surface WGPUSurface, handle to a presentable surface.
-* \param[in] w The width of the surface.
-* \param[in] h The height of the surface.
-* \param[in] device WGPUDevice, a desired handle for the wgpu device. If it is @c nullptr, ThorVG will assign an appropriate device internally.
+* @param[in] device WGPUDevice, a desired handle for the wgpu device. If it is @c nullptr, ThorVG will assign an appropriate device internally.
+* @param[in] instance WGPUInstance, context for all other wgpu objects.
+* @param[in] target Either WGPUSurface or WGPUTexture, serving as handles to a presentable surface or texture
+* @param[in] w The width of the target.
+* @param[in] h The height of the target.
+* @param[in] type 0: surface, 1: texture are used as pesentable target
 *
 * \return Tvg_Result enumeration.
 * \retval TVG_RESULT_INSUFFICIENT_CONDITION if the canvas is performing rendering. Please ensure the canvas is synced.
@@ -572,7 +573,7 @@ TVG_API Tvg_Canvas* tvg_wgcanvas_create(void);
 *
 * \note Experimental API
 */
-TVG_API Tvg_Result tvg_wgcanvas_set_target(Tvg_Canvas* canvas, void* instance, void* surface, uint32_t w, uint32_t h, void* device);
+TVG_API Tvg_Result tvg_wgcanvas_set_target(Tvg_Canvas* canvas, void* device, void* instance, void* target, uint32_t w, uint32_t h, int type = 0);
 
 /** \} */   // end defgroup ThorVGCapi_WgCanvas
 
diff --git a/src/bindings/capi/tvgCapi.cpp b/src/bindings/capi/tvgCapi.cpp
index c9457114..3ec7b601 100644
--- a/src/bindings/capi/tvgCapi.cpp
+++ b/src/bindings/capi/tvgCapi.cpp
@@ -109,10 +109,10 @@ TVG_API Tvg_Result tvg_glcanvas_set_target(Tvg_Canvas* canvas, int32_t id, uint3
 }
 
 
-TVG_API Tvg_Result tvg_wgcanvas_set_target(Tvg_Canvas* canvas, void* instance, void* surface, uint32_t w, uint32_t h, void* device)
+TVG_API Tvg_Result tvg_wgcanvas_set_target(Tvg_Canvas* canvas, void* device, void* instance, void* target, uint32_t w, uint32_t h, int type)
 {
     if (!canvas) return TVG_RESULT_INVALID_ARGUMENT;
-    return (Tvg_Result) reinterpret_cast<WgCanvas*>(canvas)->target(instance, surface, w, h, device);
+    return (Tvg_Result) reinterpret_cast<WgCanvas*>(canvas)->target(device, instance, target, w, h, type);
 }
 
 
diff --git a/src/bindings/wasm/tvgWasmLottieAnimation.cpp b/src/bindings/wasm/tvgWasmLottieAnimation.cpp
index e0447ed6..867aa01d 100644
--- a/src/bindings/wasm/tvgWasmLottieAnimation.cpp
+++ b/src/bindings/wasm/tvgWasmLottieAnimation.cpp
@@ -150,7 +150,7 @@ struct TvgWgEngine : TvgEngineMethod
     void resize(Canvas* canvas, int w, int h) override
     {
     #ifdef THORVG_WG_RASTER_SUPPORT
-        static_cast<WgCanvas*>(canvas)->target(instance, surface, w, h, device);
+        static_cast<WgCanvas*>(canvas)->target(device, instance, surface, w, h);
     #endif
     }
 };
diff --git a/src/renderer/tvgWgCanvas.cpp b/src/renderer/tvgWgCanvas.cpp
index d6dafbe6..a95a7bf2 100644
--- a/src/renderer/tvgWgCanvas.cpp
+++ b/src/renderer/tvgWgCanvas.cpp
@@ -52,25 +52,23 @@ WgCanvas::~WgCanvas()
 {
 #ifdef THORVG_WG_RASTER_SUPPORT
     auto renderer = static_cast<WgRenderer*>(Canvas::pImpl->renderer);
-    renderer->target(nullptr, 0, 0);
+    renderer->target(nullptr, nullptr, nullptr, 0, 0);
 #endif
 }
 
 
-Result WgCanvas::target(void* instance, void* surface, uint32_t w, uint32_t h, void* device) noexcept
+Result WgCanvas::target(void* device, void* instance, void* target, uint32_t w, uint32_t h, int type) noexcept
 {
 #ifdef THORVG_WG_RASTER_SUPPORT
     if (Canvas::pImpl->status != Status::Damaged && Canvas::pImpl->status != Status::Synced) {
         return Result::InsufficientCondition;
     }
 
-    if (!instance || !surface || (w == 0) || (h == 0)) return Result::InvalidArguments;
-
     //We know renderer type, avoid dynamic_cast for performance.
     auto renderer = static_cast<WgRenderer*>(Canvas::pImpl->renderer);
     if (!renderer) return Result::MemoryCorruption;
 
-    if (!renderer->target((WGPUInstance)instance, (WGPUSurface)surface, w, h, (WGPUDevice)device)) return Result::Unknown;
+    if (!renderer->target((WGPUDevice)device, (WGPUInstance)instance, target, w, h, type)) return Result::Unknown;
     Canvas::pImpl->vport = {0, 0, (int32_t)w, (int32_t)h};
     renderer->viewport(Canvas::pImpl->vport);
 
diff --git a/src/renderer/wg_engine/tvgWgBindGroups.cpp b/src/renderer/wg_engine/tvgWgBindGroups.cpp
index f3373fdb..cf8037a1 100755
--- a/src/renderer/wg_engine/tvgWgBindGroups.cpp
+++ b/src/renderer/wg_engine/tvgWgBindGroups.cpp
@@ -140,6 +140,12 @@ void WgBindGroupLayouts::releaseBindGroup(WGPUBindGroup& bindGroup)
     bindGroup = nullptr;
 }
 
+void WgBindGroupLayouts::releaseBindGroupLayout(WGPUBindGroupLayout& bindGroupLayout)
+{
+    if (bindGroupLayout) wgpuBindGroupLayoutRelease(bindGroupLayout);
+    bindGroupLayout = nullptr;
+}
+
 
 void WgBindGroupLayouts::initialize(WgContext& context)
 {
@@ -262,14 +268,15 @@ void WgBindGroupLayouts::initialize(WgContext& context)
 
 void WgBindGroupLayouts::release(WgContext& context)
 {
-    wgpuBindGroupLayoutRelease(layoutBuffer3Un);
-    wgpuBindGroupLayoutRelease(layoutBuffer2Un);
-    wgpuBindGroupLayoutRelease(layoutBuffer1Un);
-    wgpuBindGroupLayoutRelease(layoutTexStrorage3RO);
-    wgpuBindGroupLayoutRelease(layoutTexStrorage2RO);
-    wgpuBindGroupLayoutRelease(layoutTexStrorage1RO);
-    wgpuBindGroupLayoutRelease(layoutTexStrorage1WO);
-    wgpuBindGroupLayoutRelease(layoutTexSampledBuff1Un);
-    wgpuBindGroupLayoutRelease(layoutTexSampled);
+    releaseBindGroupLayout(layoutBuffer3Un);
+    releaseBindGroupLayout(layoutBuffer2Un);
+    releaseBindGroupLayout(layoutBuffer1Un);
+    releaseBindGroupLayout(layoutTexStrorage3RO);
+    releaseBindGroupLayout(layoutTexStrorage2RO);
+    releaseBindGroupLayout(layoutTexStrorage1RO);
+    releaseBindGroupLayout(layoutTexStrorage1WO);
+    releaseBindGroupLayout(layoutTexSampledBuff1Un);
+    releaseBindGroupLayout(layoutTexSampledBuff2Un);
+    releaseBindGroupLayout(layoutTexSampled);
     device = nullptr;
 }
diff --git a/src/renderer/wg_engine/tvgWgBindGroups.h b/src/renderer/wg_engine/tvgWgBindGroups.h
index ee587b60..b41bb9eb 100755
--- a/src/renderer/wg_engine/tvgWgBindGroups.h
+++ b/src/renderer/wg_engine/tvgWgBindGroups.h
@@ -51,6 +51,7 @@ public:
     WGPUBindGroup createBindGroupBuffer2Un(WGPUBuffer buff0, WGPUBuffer buff1);
     WGPUBindGroup createBindGroupBuffer3Un(WGPUBuffer buff0, WGPUBuffer buff1, WGPUBuffer buff2);
     void releaseBindGroup(WGPUBindGroup& bindGroup);
+    void releaseBindGroupLayout(WGPUBindGroupLayout& bindGroupLayout);
 public:
     void initialize(WgContext& context);
     void release(WgContext& context);
diff --git a/src/renderer/wg_engine/tvgWgCommon.cpp b/src/renderer/wg_engine/tvgWgCommon.cpp
index 5077611b..1bd6f1a6 100755
--- a/src/renderer/wg_engine/tvgWgCommon.cpp
+++ b/src/renderer/wg_engine/tvgWgCommon.cpp
@@ -260,7 +260,7 @@ void WgContext::releaseBuffer(WGPUBuffer& buffer)
     }
 }
 
-void WgContext::releaseQueue(WGPUQueue queue)
+void WgContext::releaseQueue(WGPUQueue& queue)
 {
     if (queue) {
         wgpuQueueRelease(queue);
diff --git a/src/renderer/wg_engine/tvgWgCommon.h b/src/renderer/wg_engine/tvgWgCommon.h
index a6c6cd19..ace62e94 100755
--- a/src/renderer/wg_engine/tvgWgCommon.h
+++ b/src/renderer/wg_engine/tvgWgCommon.h
@@ -63,7 +63,7 @@ struct WgContext {
     void releaseTextureView(WGPUTextureView& textureView);
     void releaseTexture(WGPUTexture& texture);
     void releaseSampler(WGPUSampler& sampler);
-    void releaseQueue(WGPUQueue queue);
+    void releaseQueue(WGPUQueue& queue);
 
     // create buffer objects (return true, if buffer handle was changed)
     bool allocateBufferUniform(WGPUBuffer& buffer, const void* data, uint64_t size);
diff --git a/src/renderer/wg_engine/tvgWgRenderData.cpp b/src/renderer/wg_engine/tvgWgRenderData.cpp
index 8f96494b..9fbddd68 100755
--- a/src/renderer/wg_engine/tvgWgRenderData.cpp
+++ b/src/renderer/wg_engine/tvgWgRenderData.cpp
@@ -165,7 +165,8 @@ void WgMeshDataPool::release(WgContext& context)
     mList.clear();
 }
 
-WgMeshDataPool* WgMeshDataGroup::gMeshDataPool = nullptr;
+WgMeshDataPool gMeshDataPoolInstance;
+WgMeshDataPool* WgMeshDataPool::gMeshDataPool = &gMeshDataPoolInstance;
 
 //***********************************************************************
 // WgMeshDataGroup
@@ -174,7 +175,7 @@ WgMeshDataPool* WgMeshDataGroup::gMeshDataPool = nullptr;
 void WgMeshDataGroup::append(WgContext& context, const WgVertexBuffer& vertexBuffer)
 {
     assert(vertexBuffer.vcount >= 3);
-    meshes.push(gMeshDataPool->allocate(context));
+    meshes.push(WgMeshDataPool::gMeshDataPool->allocate(context));
     meshes.last()->update(context, vertexBuffer);
 }
 
@@ -182,14 +183,14 @@ void WgMeshDataGroup::append(WgContext& context, const WgVertexBuffer& vertexBuf
 void WgMeshDataGroup::append(WgContext& context, const WgVertexBufferInd& vertexBufferInd)
 {
     assert(vertexBufferInd.vcount >= 3);
-    meshes.push(gMeshDataPool->allocate(context));
+    meshes.push(WgMeshDataPool::gMeshDataPool->allocate(context));
     meshes.last()->update(context, vertexBufferInd);
 }
 
 
 void WgMeshDataGroup::append(WgContext& context, const Point pmin, const Point pmax)
 {
-    meshes.push(gMeshDataPool->allocate(context));
+    meshes.push(WgMeshDataPool::gMeshDataPool->allocate(context));
     meshes.last()->bbox(context, pmin, pmax);
 }
 
@@ -197,7 +198,7 @@ void WgMeshDataGroup::append(WgContext& context, const Point pmin, const Point p
 void WgMeshDataGroup::release(WgContext& context)
 {
     for (uint32_t i = 0; i < meshes.count; i++)
-        gMeshDataPool->free(context, meshes[i]);
+        WgMeshDataPool::gMeshDataPool->free(context, meshes[i]);
     meshes.clear();
 };
 
@@ -298,8 +299,6 @@ void WgRenderSettings::release(WgContext& context)
 // WgRenderDataPaint
 //***********************************************************************
 
-WgVertexBufferInd* WgRenderDataPaint::gStrokesGenerator = nullptr;
-
 void WgRenderDataPaint::release(WgContext& context)
 {
     context.pipelines->layouts.releaseBindGroup(bindGroupPaint);
@@ -455,7 +454,8 @@ void WgRenderDataShape::updateMeshes(WgContext &context, const RenderShape &rsha
 void WgRenderDataShape::proceedStrokes(WgContext context, const RenderStroke* rstroke, float tbeg, float tend, const WgVertexBuffer& buff)
 {
     assert(rstroke);
-    gStrokesGenerator->reset(buff.tscale);
+    static WgVertexBufferInd strokesGenerator;
+    strokesGenerator.reset(buff.tscale);
     // trim -> dash -> stroke
     if ((tbeg != 0.0f) || (tend != 1.0f)) {
         if (tbeg == tend) return;
@@ -464,17 +464,17 @@ void WgRenderDataShape::proceedStrokes(WgContext context, const RenderStroke* rs
         trimed_buff.trim(buff, tbeg, tend);
         trimed_buff.updateDistances();
         // trim ->dash -> stroke
-        if (rstroke->dashPattern) gStrokesGenerator->appendStrokesDashed(trimed_buff, rstroke);
+        if (rstroke->dashPattern) strokesGenerator.appendStrokesDashed(trimed_buff, rstroke);
         // trim -> stroke
-        else gStrokesGenerator->appendStrokes(trimed_buff, rstroke);
+        else strokesGenerator.appendStrokes(trimed_buff, rstroke);
     } else
     // dash -> stroke
     if (rstroke->dashPattern) {
-        gStrokesGenerator->appendStrokesDashed(buff, rstroke);
+        strokesGenerator.appendStrokesDashed(buff, rstroke);
     // stroke
     } else
-        gStrokesGenerator->appendStrokes(buff, rstroke);
-    appendStroke(context, *gStrokesGenerator);
+        strokesGenerator.appendStrokes(buff, rstroke);
+    appendStroke(context, strokesGenerator);
 }
 
 
diff --git a/src/renderer/wg_engine/tvgWgRenderData.h b/src/renderer/wg_engine/tvgWgRenderData.h
index 79ee1ece..1b091688 100755
--- a/src/renderer/wg_engine/tvgWgRenderData.h
+++ b/src/renderer/wg_engine/tvgWgRenderData.h
@@ -50,14 +50,13 @@ private:
     Array<WgMeshData*> mPool;
     Array<WgMeshData*> mList;
 public:
+    static WgMeshDataPool* gMeshDataPool;
     WgMeshData* allocate(WgContext& context);
     void free(WgContext& context, WgMeshData* meshData);
     void release(WgContext& context);
 };
 
 struct WgMeshDataGroup {
-    static WgMeshDataPool* gMeshDataPool;
-
     Array<WgMeshData*> meshes{};
     
     void append(WgContext& context, const WgVertexBuffer& vertexBuffer);
@@ -96,9 +95,6 @@ struct WgRenderSettings
 
 struct WgRenderDataPaint
 {
-    // global strokes generator. single instance
-    static WgVertexBufferInd* gStrokesGenerator;
-
     WGPUBuffer bufferModelMat{};
     WGPUBuffer bufferBlendSettings{};
     WGPUBindGroup bindGroupPaint{};
diff --git a/src/renderer/wg_engine/tvgWgRenderer.cpp b/src/renderer/wg_engine/tvgWgRenderer.cpp
index cb4d02fd..c8e6342d 100755
--- a/src/renderer/wg_engine/tvgWgRenderer.cpp
+++ b/src/renderer/wg_engine/tvgWgRenderer.cpp
@@ -30,47 +30,45 @@ WgRenderer::WgRenderer()
 WgRenderer::~WgRenderer()
 {
     release();
-    mContext.release();
 }
 
 
 void WgRenderer::initialize()
 {
-    mPipelines.initialize(mContext);
-    WgMeshDataGroup::gMeshDataPool = new WgMeshDataPool();
-    WgRenderDataShape::gStrokesGenerator = new WgVertexBufferInd();
 }
 
 
 void WgRenderer::release()
 {
+    // check for general context availability
+    if (!mContext.queue) return;
+
+    // dispose stored objects
     disposeObjects();
-    mStorageRoot.release(mContext);
-    mRenderStoragePool.release(mContext);
-    mRenderDataShapePool.release(mContext);
-    delete WgRenderDataShape::gStrokesGenerator;
-    WgMeshDataGroup::gMeshDataPool->release(mContext);
-    delete WgMeshDataGroup::gMeshDataPool;
+
+    // clear rendering tree stacks
     mCompositorStack.clear();
     mRenderStorageStack.clear();
+    mRenderStoragePool.release(mContext);
+    mRenderDataShapePool.release(mContext);
+    WgMeshDataPool::gMeshDataPool->release(mContext);
+    mStorageRoot.release(mContext);
+
+    // release context handles
+    mCompositor.release(mContext);
     mPipelines.release(mContext);
-    if (gpuOwner) {
-        if (device) wgpuDeviceRelease(device);
-        device = nullptr;
-        if (adapter) wgpuAdapterRelease(adapter);
-        adapter = nullptr;
-        gpuOwner = false;
-    }
-    releaseSurfaceTexture();
+    mContext.release();
+
+    // release gpu handles
+    clearTargets();
+    releaseDevice();
 }
 
 
 void WgRenderer::disposeObjects()
 {
-    if (mDisposeRenderDatas.count == 0) return;
-
-    for (auto p = mDisposeRenderDatas.begin(); p < mDisposeRenderDatas.end(); p++) {
-        auto renderData = (WgRenderDataPaint*)(*p);
+    for (uint32_t i = 0; i < mDisposeRenderDatas.count; i++) {
+        WgRenderDataPaint* renderData = (WgRenderDataPaint*)mDisposeRenderDatas[i];
         if (renderData->type() == Type::Shape) {
             mRenderDataShapePool.free(mContext, (WgRenderDataShape*)renderData);
         } else {
@@ -260,20 +258,26 @@ void WgRenderer::releaseSurfaceTexture()
 bool WgRenderer::sync()
 {
     disposeObjects();
-    if (!surface) return false;
 
-    releaseSurfaceTexture();
+    // if texture buffer used
+    WGPUTexture dstTexture = targetTexture;
+    if (surface) {
+        releaseSurfaceTexture();
+        wgpuSurfaceGetCurrentTexture(surface, &surfaceTexture);
+        dstTexture = surfaceTexture.texture;
+    }
+    // there is no external dest buffer
+    if (!dstTexture) return false;
 
-    wgpuSurfaceGetCurrentTexture(surface, &surfaceTexture);
-
-    WGPUTextureView dstView = mContext.createTextureView(surfaceTexture.texture);
+    // get external dest buffer
+    WGPUTextureView dstTextureView = mContext.createTextureView(dstTexture);
 
     // create command encoder
     const WGPUCommandEncoderDescriptor commandEncoderDesc{};
     WGPUCommandEncoder commandEncoder = wgpuDeviceCreateCommandEncoder(mContext.device, &commandEncoderDesc);
 
     // show root offscreen buffer
-    mCompositor.blit(mContext, commandEncoder, &mStorageRoot, dstView);
+    mCompositor.blit(mContext, commandEncoder, &mStorageRoot, dstTextureView);
 
     // release command encoder
     const WGPUCommandBufferDescriptor commandBufferDesc{};
@@ -281,68 +285,109 @@ bool WgRenderer::sync()
     wgpuQueueSubmit(mContext.queue, 1, &commandsBuffer);
     wgpuCommandBufferRelease(commandsBuffer);
     wgpuCommandEncoderRelease(commandEncoder);
-    mContext.releaseTextureView(dstView);
 
+    // release dest buffer view
+    mContext.releaseTextureView(dstTextureView);
     return true;
 }
 
 
-// target for native window handle
-bool WgRenderer::target(WGPUInstance instance, WGPUSurface surface, uint32_t w, uint32_t h, WGPUDevice device)
+// render target handle
+bool WgRenderer::target(WGPUDevice device, WGPUInstance instance, void* target, uint32_t width, uint32_t height, int type)
 {
-    gpuOwner = false;
+    // release existing handles
+    release();
+
+    // can not initialize renderer, give up
+    if (!instance || !target || !width || !height) return false;
+
+    // store or regest gpu device
     this->device = device;
-    if (!this->device) {
-        // request adapter
-        const WGPURequestAdapterOptions requestAdapterOptions { .nextInChain = nullptr, .compatibleSurface = surface, .powerPreference = WGPUPowerPreference_HighPerformance, .forceFallbackAdapter = false };
-        auto onAdapterRequestEnded = [](WGPURequestAdapterStatus status, WGPUAdapter adapter, char const * message, void * pUserData) { *((WGPUAdapter*)pUserData) = adapter; };
-        wgpuInstanceRequestAdapter(instance, &requestAdapterOptions, onAdapterRequestEnded, &this->adapter);
+    if (!this->device)
+        reguestDevice(instance, (WGPUSurface)target);
 
-        // get adapter and surface properties
-        WGPUFeatureName featureNames[32]{};
-        size_t featuresCount = wgpuAdapterEnumerateFeatures(this->adapter, featureNames);
-
-        // request device
-        const WGPUDeviceDescriptor deviceDesc { .nextInChain = nullptr, .label = "The device", .requiredFeatureCount = featuresCount, .requiredFeatures = featureNames };
-        auto onDeviceRequestEnded = [](WGPURequestDeviceStatus status, WGPUDevice device, char const * message, void * pUserData) { *((WGPUDevice*)pUserData) = device; };
-        wgpuAdapterRequestDevice(this->adapter, &deviceDesc, onDeviceRequestEnded, &this->device);
-        gpuOwner = true;
-    }
+    // store target properties
+    mTargetSurface.stride = width;
+    mTargetSurface.w = width;
+    mTargetSurface.h = height;
 
+    // initialize rendering context
     mContext.initialize(instance, this->device);
-    initialize();
-    target(surface, w, h);
-    mRenderStoragePool.initialize(mContext, w, h);
-    mStorageRoot.initialize(mContext, w, h);
-    mCompositor.initialize(mContext, w, h);
+    mPipelines.initialize(mContext);
+
+    // initialize render tree instances
+    mRenderStoragePool.initialize(mContext, width, height);
+    mStorageRoot.initialize(mContext, width, height);
+    mCompositor.initialize(mContext, width, height);
+
+    // configure surface (must be called after context creation)
+    if (type == 0) {
+        surface = (WGPUSurface)target;
+        surfaceConfigure(surface, mContext, width, height);
+    } else targetTexture = (WGPUTexture)target;
     return true;
 }
 
-bool WgRenderer::target(WGPUSurface surface, uint32_t w, uint32_t h) {
+
+void WgRenderer::reguestDevice(WGPUInstance instance, WGPUSurface surface)
+{
+    // request adapter
+    const WGPURequestAdapterOptions requestAdapterOptions { .compatibleSurface = surface, .powerPreference = WGPUPowerPreference_HighPerformance };
+    auto onAdapterRequestEnded = [](WGPURequestAdapterStatus status, WGPUAdapter adapter, char const * message, void * pUserData) { *((WGPUAdapter*)pUserData) = adapter; };
+    wgpuInstanceRequestAdapter(instance, &requestAdapterOptions, onAdapterRequestEnded, &this->adapter);
+
+    // get adapter and surface properties
+    WGPUFeatureName featureNames[32]{};
+    size_t featuresCount = wgpuAdapterEnumerateFeatures(this->adapter, featureNames);
+
+    // request device
+    const WGPUDeviceDescriptor deviceDesc { .label = "The owned device", .requiredFeatureCount = featuresCount, .requiredFeatures = featureNames };
+    auto onDeviceRequestEnded = [](WGPURequestDeviceStatus status, WGPUDevice device, char const * message, void * pUserData) { *((WGPUDevice*)pUserData) = device; };
+    wgpuAdapterRequestDevice(this->adapter, &deviceDesc, onDeviceRequestEnded, &this->device);
+    gpuOwner = true;
+}
+
+
+void WgRenderer::releaseDevice()
+{
+    if (!gpuOwner) return;
+    wgpuDeviceRelease(device);
+    wgpuDeviceDestroy(device);
+    wgpuAdapterRelease(adapter);
+    device = nullptr;
+    adapter = nullptr;
+    gpuOwner = false;
+}
+
+
+void WgRenderer::clearTargets() {
+    releaseSurfaceTexture();
+    targetTexture = nullptr;
+    surface = nullptr;
+}
+
+
+bool WgRenderer::surfaceConfigure(WGPUSurface surface, WgContext& context, uint32_t width, uint32_t height)
+{
     // store target surface properties
     this->surface = surface;
-    mTargetSurface.stride = w;
-    mTargetSurface.w = w;
-    mTargetSurface.h = h;
-    if (w == 0 || h == 0) return false;
-    if (!surface) return true;
+    if (width == 0 || height == 0 || !surface) return false;
 
+    // setup surface configuration
     WGPUSurfaceConfiguration surfaceConfiguration {
-        .device = mContext.device,
-        .format = mContext.preferredFormat,
+        .device = context.device,
+        .format = context.preferredFormat,
         .usage = WGPUTextureUsage_RenderAttachment,
     #ifdef __EMSCRIPTEN__
         .alphaMode = WGPUCompositeAlphaMode_Premultiplied,
     #endif
-        .width = w, .height = h,
-    #ifdef __EMSCRIPTEN__
+        .width = width,
+        .height = height,
+    #ifndef __EMSCRIPTEN__
         .presentMode = WGPUPresentMode_Fifo,
-    #else
-        .presentMode = WGPUPresentMode_Immediate
     #endif
     };
     wgpuSurfaceConfigure(surface, &surfaceConfiguration);
-
     return true;
 }
 
@@ -369,7 +414,7 @@ bool WgRenderer::beginComposite(RenderCompositor* cmp, MaskMethod method, uint8_
     mRenderStorageStack.push(storage);
     // begin newly added render pass
     WGPUColor color{};
-    if ((method == MaskMethod::None) && (opacity != 255)) color = {1.0f, 1.0f, 1.0f, 0.0f};
+    if ((method == MaskMethod::None) && (opacity != 255)) color = { 1.0, 1.0, 1.0, 0.0 };
     mCompositor.beginRenderPass(mCommandEncoder, mRenderStorageStack.last(), true, color);
     return true;
 }
diff --git a/src/renderer/wg_engine/tvgWgRenderer.h b/src/renderer/wg_engine/tvgWgRenderer.h
index 9faa7317..6bb37819 100755
--- a/src/renderer/wg_engine/tvgWgRenderer.h
+++ b/src/renderer/wg_engine/tvgWgRenderer.h
@@ -45,8 +45,7 @@ public:
     bool clear() override;
     bool sync() override;
 
-    bool target(WGPUInstance instance, WGPUSurface surface, uint32_t w, uint32_t h, WGPUDevice device);
-    bool target(WGPUSurface surface, uint32_t w, uint32_t h);
+    bool target(WGPUDevice device, WGPUInstance instance, void* target, uint32_t width, uint32_t height, int type = 0);
 
     RenderCompositor* target(const RenderRegion& region, ColorSpace cs) override;
     bool beginComposite(RenderCompositor* cmp, MaskMethod method, uint8_t opacity) override;
@@ -59,8 +58,6 @@ public:
     static bool init(uint32_t threads);
     static bool term();
 
-    WGPUSurface surface{}; // external handle
-
 private:
     WgRenderer();
     ~WgRenderer();
@@ -69,28 +66,37 @@ private:
     void disposeObjects();
     void releaseSurfaceTexture();
 
-    WGPUSurfaceTexture surfaceTexture{};
+    void reguestDevice(WGPUInstance instance, WGPUSurface surface);
+    void releaseDevice();
+    void clearTargets();
+    bool surfaceConfigure(WGPUSurface surface, WgContext& context, uint32_t width, uint32_t height);
 
-    WGPUCommandEncoder mCommandEncoder{};
-    WgRenderDataShapePool mRenderDataShapePool;
-
-    // render tree stacks
+    // render tree stacks and pools
     WgRenderStorage mStorageRoot;
     Array<WgCompose*> mCompositorStack;
     Array<WgRenderStorage*> mRenderStorageStack;
     WgRenderStoragePool mRenderStoragePool;
+    WgRenderDataShapePool mRenderDataShapePool;
 
+    // rendering context
     WgContext mContext;
     WgPipelines mPipelines;
     WgCompositor mCompositor;
 
+    // rendering states
     RenderSurface mTargetSurface;
     BlendMethod mBlendMethod{};
     RenderRegion mViewport{};
 
+    // disposable data list
     Array<RenderData> mDisposeRenderDatas{};
     Key mDisposeKey{};
 
+    // gpu handles
+    WGPUCommandEncoder mCommandEncoder{};
+    WGPUTexture targetTexture{}; // external handle
+    WGPUSurfaceTexture surfaceTexture{};
+    WGPUSurface surface{};  // external handle
     WGPUAdapter adapter{};
     WGPUDevice device{};
     bool gpuOwner{};