diff --git a/backend/internal/service/canvas_share.go b/backend/internal/service/canvas_share.go
index 10c71aa3e..0fdc88ac7 100644
--- a/backend/internal/service/canvas_share.go
+++ b/backend/internal/service/canvas_share.go
@@ -36,7 +36,7 @@ type PublicCanvasShare struct {
 
 var publicCanvasMetadataKeys = map[string]bool{
 	"content": true, "composerContent": true, "prompt": true, "status": true, "fontSize": true,
-	"generationMode": true, "generationType": true, "model": true, "size": true, "quality": true,
+	"generationMode": true, "generationType": true, "model": true, "size": true, "quality": true, "transparentBackground": true,
 	"count": true, "seconds": true, "vquality": true, "generateAudio": true, "watermark": true,
 	"audioVoice": true, "audioFormat": true, "audioSpeed": true, "audioInstructions": true,
 	"naturalWidth": true, "naturalHeight": true, "freeResize": true, "isBatchRoot": true,
diff --git a/backend/internal/service/provider.go b/backend/internal/service/provider.go
index 1bd99c47f..596af7742 100644
--- a/backend/internal/service/provider.go
+++ b/backend/internal/service/provider.go
@@ -31,24 +31,25 @@ type canvasGenerationInput struct {
 }
 
 type providerConfig struct {
-	ChannelID          string `json:"channelId"`
-	APIFormat          string `json:"apiFormat"`
-	InterfaceType      string `json:"interfaceType"`
-	BaseURL            string `json:"baseUrl"`
-	APIKey             string `json:"apiKey"`
-	Model              string `json:"model"`
-	Size               string `json:"size"`
-	Quality            string `json:"quality"`
-	Count              string `json:"count"`
-	VideoSeconds       string `json:"videoSeconds"`
-	VQuality           string `json:"vquality"`
-	VideoGenerateAudio string `json:"videoGenerateAudio"`
-	VideoWatermark     string `json:"videoWatermark"`
-	AudioVoice         string `json:"audioVoice"`
-	AudioFormat        string `json:"audioFormat"`
-	AudioSpeed         string `json:"audioSpeed"`
-	AudioInstructions  string `json:"audioInstructions"`
-	SystemPrompt       string `json:"systemPrompt"`
+	ChannelID             string `json:"channelId"`
+	APIFormat             string `json:"apiFormat"`
+	InterfaceType         string `json:"interfaceType"`
+	BaseURL               string `json:"baseUrl"`
+	APIKey                string `json:"apiKey"`
+	Model                 string `json:"model"`
+	Size                  string `json:"size"`
+	Quality               string `json:"quality"`
+	TransparentBackground string `json:"transparentBackground"`
+	Count                 string `json:"count"`
+	VideoSeconds          string `json:"videoSeconds"`
+	VQuality              string `json:"vquality"`
+	VideoGenerateAudio    string `json:"videoGenerateAudio"`
+	VideoWatermark        string `json:"videoWatermark"`
+	AudioVoice            string `json:"audioVoice"`
+	AudioFormat           string `json:"audioFormat"`
+	AudioSpeed            string `json:"audioSpeed"`
+	AudioInstructions     string `json:"audioInstructions"`
+	SystemPrompt          string `json:"systemPrompt"`
 }
 
 const providerHTTPTimeout = 5 * time.Minute
@@ -325,6 +326,9 @@ func systemChannelIDFromBaseURL(baseURL string) string {
 }
 
 func runImageTask(ctx context.Context, input canvasGenerationInput) (map[string]interface{}, error) {
+	if input.Config.TransparentBackground == "true" && !supportsTransparentImageBackground(input.Config) {
+		return nil, errors.New("当前模型不支持透明背景，请关闭透明背景或切换至 GPT Image 1 系列模型")
+	}
 	var payload imageResponse
 	if len(input.ReferenceImages) > 0 || input.Mask != nil {
 		body := &bytes.Buffer{}
@@ -334,6 +338,9 @@ func runImageTask(ctx context.Context, input canvasGenerationInput) (map[string]
 		writeField(writer, "n", "1")
 		writeField(writer, "response_format", "b64_json")
 		writeField(writer, "output_format", "png")
+		if input.Config.TransparentBackground == "true" {
+			writeField(writer, "background", "transparent")
+		}
 		if input.Config.Quality != "" {
 			writeField(writer, "quality", normalizeImageQuality(input.Config.Quality))
 		}
@@ -364,6 +371,9 @@ func runImageTask(ctx context.Context, input canvasGenerationInput) (map[string]
 			"response_format": "b64_json",
 			"output_format":   "png",
 		}
+		if input.Config.TransparentBackground == "true" {
+			body["background"] = "transparent"
+		}
 		if input.Config.Quality != "" {
 			body["quality"] = normalizeImageQuality(input.Config.Quality)
 		}
@@ -1762,6 +1772,12 @@ func isArkPlanVideoConfig(config providerConfig) bool {
 	return strings.Contains(strings.ToLower(config.BaseURL), "/api/plan/v3")
 }
 
+// 透明背景必须按模型能力显式开放，避免上游静默忽略参数后返回不透明图片。
+func supportsTransparentImageBackground(config providerConfig) bool {
+	model := strings.ToLower(strings.TrimSpace(config.Model))
+	return strings.EqualFold(strings.TrimSpace(config.APIFormat), "openai") && (model == "gpt-image-1" || strings.HasPrefix(model, "gpt-image-1-") || strings.HasPrefix(model, "gpt-image-1."))
+}
+
 func normalizeImageQuality(value string) string {
 	switch strings.ToLower(strings.TrimSpace(value)) {
 	case "1k":
diff --git a/web/src/components/canvas/canvas-assistant-panel.tsx b/web/src/components/canvas/canvas-assistant-panel.tsx
index 8a8883af9..7d556ac5e 100644
--- a/web/src/components/canvas/canvas-assistant-panel.tsx
+++ b/web/src/components/canvas/canvas-assistant-panel.tsx
@@ -36,6 +36,7 @@ const GENERATION_OPTION_PROPERTIES = {
     model: { type: "string" },
     size: { type: "string" },
     quality: { type: "string" },
+    transparentBackground: { type: "string", enum: ["true", "false"] },
     count: { type: "number" },
     seconds: { type: "string" },
     vquality: { type: "string" },
@@ -1239,6 +1240,7 @@ function configNodeOp(id: string, input: Record<string, unknown>, x: number, y:
             model: resolveGenerationModel(config, mode, stringOptional(input.model)),
             size: stringOptional(input.size) || config.size,
             quality: stringOptional(input.quality) || config.quality,
+            transparentBackground: stringOptional(input.transparentBackground) || config.transparentBackground,
             count: numberOptional(input.count) ?? generationCount(mode === "image" ? config.canvasImageCount || config.count : config.count),
             seconds: stringOptional(input.seconds) || config.videoSeconds,
             vquality: stringOptional(input.vquality) || config.vquality,
@@ -1538,6 +1540,7 @@ function backendAgentProviderConfig(config: ReturnType<typeof resolveModelReques
         model: config.model,
         size: config.size,
         quality: config.quality,
+        transparentBackground: config.transparentBackground,
         count: config.count,
         videoSeconds: config.videoSeconds,
         vquality: config.vquality,
diff --git a/web/src/components/canvas/canvas-config-node-panel.tsx b/web/src/components/canvas/canvas-config-node-panel.tsx
index 8f8e2e5cb..22f0a32af 100644
--- a/web/src/components/canvas/canvas-config-node-panel.tsx
+++ b/web/src/components/canvas/canvas-config-node-panel.tsx
@@ -214,6 +214,7 @@ function buildNodeConfig(globalConfig: AiConfig, node: CanvasNodeData, mode: Can
         model,
         quality: node.metadata?.quality || globalConfig.quality || defaultConfig.quality,
         size: node.metadata?.size || globalConfig.size || defaultConfig.size,
+        transparentBackground: (node.metadata?.transparentBackground || globalConfig.transparentBackground) === "true" ? "true" : "false",
         videoSeconds: normalizeVideoDuration(node.metadata?.seconds || globalConfig.videoSeconds || defaultConfig.videoSeconds),
         vquality: normalizeVideoResolution(node.metadata?.vquality || globalConfig.vquality || defaultConfig.vquality),
         videoGenerateAudio: node.metadata?.generateAudio || globalConfig.videoGenerateAudio || defaultConfig.videoGenerateAudio,
diff --git a/web/src/components/canvas/canvas-image-settings-popover.tsx b/web/src/components/canvas/canvas-image-settings-popover.tsx
index 7df26345f..a0cc358dc 100644
--- a/web/src/components/canvas/canvas-image-settings-popover.tsx
+++ b/web/src/components/canvas/canvas-image-settings-popover.tsx
@@ -29,7 +29,8 @@ export function CanvasImageSettingsPopover({ config, onConfigChange, onOpenChang
     const quality = config.quality || "auto";
     const count = Math.max(1, Math.min(15, Math.floor(Math.abs(Number(config.count)) || 1)));
     const activeSize = config.size || "auto";
-    const summary = showCount ? `${imageQualityLabel(quality)} · ${imageSizeLabel(activeSize)} · ${count} 张` : `${imageQualityLabel(quality)} · ${imageSizeLabel(activeSize)}`;
+    const transparentLabel = config.transparentBackground === "true" ? " · 透明" : "";
+    const summary = showCount ? `${imageQualityLabel(quality)} · ${imageSizeLabel(activeSize)} · ${count} 张${transparentLabel}` : `${imageQualityLabel(quality)} · ${imageSizeLabel(activeSize)}${transparentLabel}`;
     const updateOpen = (nextOpen: boolean) => {
         setOpen(nextOpen);
         onOpenChange?.(nextOpen);
diff --git a/web/src/components/canvas/canvas-node-prompt-panel.tsx b/web/src/components/canvas/canvas-node-prompt-panel.tsx
index 402b9401a..4cdd18ace 100644
--- a/web/src/components/canvas/canvas-node-prompt-panel.tsx
+++ b/web/src/components/canvas/canvas-node-prompt-panel.tsx
@@ -377,6 +377,7 @@ function buildNodeConfig(globalConfig: AiConfig, node: CanvasNodeData, mode: Can
         model,
         quality: node.metadata?.quality || globalConfig.quality || defaultConfig.quality,
         size: node.metadata?.size || globalConfig.size || defaultConfig.size,
+        transparentBackground: (node.metadata?.transparentBackground || globalConfig.transparentBackground) === "true" ? "true" : "false",
         videoSeconds: normalizeVideoDuration(node.metadata?.seconds || globalConfig.videoSeconds || defaultConfig.videoSeconds),
         vquality: normalizeVideoResolution(node.metadata?.vquality || globalConfig.vquality || defaultConfig.vquality),
         videoGenerateAudio: node.metadata?.generateAudio || globalConfig.videoGenerateAudio || defaultConfig.videoGenerateAudio,
diff --git a/web/src/components/image-settings-panel.tsx b/web/src/components/image-settings-panel.tsx
index 7a510200c..12f17d14e 100644
--- a/web/src/components/image-settings-panel.tsx
+++ b/web/src/components/image-settings-panel.tsx
@@ -2,7 +2,7 @@ import { type ReactNode, useState } from "react";
 import { ConfigProvider, Switch } from "antd";
 
 import { type CanvasTheme } from "@/lib/canvas-theme";
-import type { AiConfig } from "@/stores/use-config-store";
+import { supportsTransparentImageBackground, type AiConfig } from "@/stores/use-config-store";
 
 const qualityOptions = [
     { value: "auto", label: "自动" },
@@ -30,7 +30,7 @@ const aspectOptions = [
 
 type ImageSettingsPanelProps = {
     config: AiConfig;
-    onConfigChange: (key: "quality" | "size" | "count", value: string) => void;
+    onConfigChange: (key: "quality" | "size" | "transparentBackground" | "count", value: string) => void;
     theme: CanvasTheme;
     showTitle?: boolean;
     showCount?: boolean;
@@ -42,6 +42,8 @@ type ImageSettingsPanelProps = {
 export function ImageSettingsPanel({ config, onConfigChange, theme, showTitle = true, showCount = true, className = "w-[304px] space-y-3 rounded-2xl px-1 py-0.5", maxCount = 15, quickCount = 10 }: ImageSettingsPanelProps) {
     const [snapDimensionToStep, setSnapDimensionToStep] = useState(true);
     const quality = config.quality || "auto";
+    const transparentBackground = config.transparentBackground === "true";
+    const transparentBackgroundSupported = supportsTransparentImageBackground(config, config.model || config.imageModel);
     const count = Math.max(1, Math.min(maxCount, Math.floor(Math.abs(Number(config.count)) || 1)));
     const activeSize = config.size || "auto";
     const selectedAspect = aspectOptions.find((item) => (item.size || item.value) === activeSize || item.value === activeSize);
@@ -79,6 +81,22 @@ export function ImageSettingsPanel({ config, onConfigChange, theme, showTitle =
                         ))}
                     </div>
                 </div>
+                <div className="flex items-center justify-between gap-3">
+                    <div className="min-w-0">
+                        <SettingTitle color={theme.node.muted}>透明背景</SettingTitle>
+                        <div className="mt-1 text-[11px]" style={{ color: transparentBackgroundSupported ? theme.node.muted : theme.accent.danger }}>
+                            {transparentBackgroundSupported ? "输出保留 Alpha 通道的 PNG" : transparentBackground ? "当前模型不支持，生成会被阻止" : "当前模型不支持透明背景"}
+                        </div>
+                    </div>
+                    <span title={transparentBackgroundSupported ? "使用透明背景生成 PNG 图片" : "仅 GPT Image 1 系列模型支持透明背景"} onMouseDown={(event) => event.stopPropagation()}>
+                        <Switch
+                            size="small"
+                            checked={transparentBackground}
+                            disabled={!transparentBackgroundSupported && !transparentBackground}
+                            onChange={(checked) => onConfigChange("transparentBackground", checked ? "true" : "false")}
+                        />
+                    </span>
+                </div>
                 <div className="space-y-2">
                     <div className="flex items-center justify-between gap-3">
                         <SettingTitle color={theme.node.muted}>尺寸</SettingTitle>
diff --git a/web/src/lib/canvas/canvas-project-generation.ts b/web/src/lib/canvas/canvas-project-generation.ts
index b8d03c4ed..bc356e793 100644
--- a/web/src/lib/canvas/canvas-project-generation.ts
+++ b/web/src/lib/canvas/canvas-project-generation.ts
@@ -107,6 +107,7 @@ export function backendProviderConfig(config: AiConfig) {
         model: requestConfig.model,
         size: config.size,
         quality: config.quality,
+        transparentBackground: config.transparentBackground,
         count: config.count,
         videoSeconds: config.videoSeconds,
         vquality: config.vquality,
@@ -177,6 +178,7 @@ export function buildImageGenerationMetadata(type: CanvasImageGenerationType, co
         model: config.model,
         size: config.size,
         quality: config.quality,
+        transparentBackground: config.transparentBackground,
         count,
         references: references.map(referenceUrl).filter((url): url is string => Boolean(url)),
     };
@@ -338,6 +340,7 @@ export function buildGenerationConfig(config: AiConfig, node: CanvasNodeData | u
         model,
         quality: node?.metadata?.quality || config.quality || defaultConfig.quality,
         size: node?.metadata?.size || config.size || defaultConfig.size,
+        transparentBackground: (node?.metadata?.transparentBackground || config.transparentBackground) === "true" ? "true" : "false",
         videoSeconds: normalizeVideoDuration(node?.metadata?.seconds || config.videoSeconds || defaultConfig.videoSeconds),
         vquality: normalizeVideoResolution(node?.metadata?.vquality || config.vquality || defaultConfig.vquality),
         videoGenerateAudio: node?.metadata?.generateAudio || config.videoGenerateAudio || defaultConfig.videoGenerateAudio,
diff --git a/web/src/pages/canvas/project.tsx b/web/src/pages/canvas/project.tsx
index e0402e77f..aaff37610 100644
--- a/web/src/pages/canvas/project.tsx
+++ b/web/src/pages/canvas/project.tsx
@@ -1030,6 +1030,8 @@ function InfiniteCanvasPage() {
                     prompt: "",
                     model: effectiveConfig.imageModel || effectiveConfig.model,
                     size: effectiveConfig.size,
+                    quality: effectiveConfig.quality,
+                    transparentBackground: effectiveConfig.transparentBackground,
                     count: getGenerationCount(effectiveConfig.canvasImageCount || effectiveConfig.count),
                 },
             );
diff --git a/web/src/pages/canvas/use-canvas-generation-retry.ts b/web/src/pages/canvas/use-canvas-generation-retry.ts
index 38c90c26e..c35aa4ea9 100644
--- a/web/src/pages/canvas/use-canvas-generation-retry.ts
+++ b/web/src/pages/canvas/use-canvas-generation-retry.ts
@@ -62,7 +62,7 @@ export function useCanvasGenerationRetry({ projectId, activatedSkills, nodesRef,
             const retryMode: CanvasNodeGenerationMode = node.type === CanvasNodeType.Text ? "text" : node.type === CanvasNodeType.Video ? "video" : node.type === CanvasNodeType.Audio ? "audio" : "image";
             const generationConfig =
                 hasSavedImageMetadata && savedImageMetadata
-                    ? { ...effectiveConfig, model: savedImageMetadata.model || effectiveConfig.imageModel || effectiveConfig.model, quality: savedImageMetadata.quality || effectiveConfig.quality, size: savedImageMetadata.size || effectiveConfig.size, count: "1" }
+                    ? { ...effectiveConfig, model: savedImageMetadata.model || effectiveConfig.imageModel || effectiveConfig.model, quality: savedImageMetadata.quality || effectiveConfig.quality, size: savedImageMetadata.size || effectiveConfig.size, transparentBackground: (savedImageMetadata.transparentBackground || effectiveConfig.transparentBackground) === "true" ? "true" : "false", count: "1" }
                     : { ...buildGenerationConfig(effectiveConfig, sourceNode, retryMode), count: "1" };
             if (!isAiConfigReady(generationConfig, generationConfig.model)) {
                 openConfigDialog(true);
@@ -135,7 +135,7 @@ export function useCanvasGenerationRetry({ projectId, activatedSkills, nodesRef,
                 const imageConfig = NODE_DEFAULT_SIZE[CanvasNodeType.Image];
                 const imageSize = fitNodeSize(uploadedImage.width, uploadedImage.height, imageConfig.width, imageConfig.height);
                 const generationMetadata = savedImageMetadata?.generationType
-                    ? { generationType: savedImageMetadata.generationType, model: generationConfig.model, size: generationConfig.size, quality: generationConfig.quality, count: savedImageMetadata.count || 1, references: savedImageMetadata.references }
+                    ? { generationType: savedImageMetadata.generationType, model: generationConfig.model, size: generationConfig.size, quality: generationConfig.quality, transparentBackground: generationConfig.transparentBackground, count: savedImageMetadata.count || 1, references: savedImageMetadata.references }
                     : buildImageGenerationMetadata(useReferenceImages ? "edit" : "generation", generationConfig, 1, retryImages);
                 setNodes((current) => current.map((item) => (item.id === node.id ? { ...item, type: CanvasNodeType.Image, width: imageSize.width, height: imageSize.height, metadata: { ...item.metadata, ...imageMetadata(uploadedImage), prompt, ...generationMetadata } } : item)));
             } catch (error) {
diff --git a/web/src/pages/canvas/use-canvas-upload.ts b/web/src/pages/canvas/use-canvas-upload.ts
index 547f900d5..15b8a8ff2 100644
--- a/web/src/pages/canvas/use-canvas-upload.ts
+++ b/web/src/pages/canvas/use-canvas-upload.ts
@@ -278,6 +278,7 @@ export function useCanvasUpload({
                     model: undefined,
                     size: undefined,
                     quality: undefined,
+                    transparentBackground: undefined,
                     count: undefined,
                     references: undefined,
                     primaryImageId: undefined,
diff --git a/web/src/pages/tasks/index.tsx b/web/src/pages/tasks/index.tsx
index c187ea7ef..c03e7f714 100644
--- a/web/src/pages/tasks/index.tsx
+++ b/web/src/pages/tasks/index.tsx
@@ -493,6 +493,7 @@ function backendProviderConfig(config: ReturnType<typeof resolveModelRequestConf
         model: config.model,
         size: config.size,
         quality: config.quality,
+        transparentBackground: config.transparentBackground,
         count: config.count,
         videoSeconds: config.videoSeconds,
         vquality: config.vquality,
diff --git a/web/src/services/api/image.ts b/web/src/services/api/image.ts
index be2af3fdf..1e408edfc 100644
--- a/web/src/services/api/image.ts
+++ b/web/src/services/api/image.ts
@@ -1,6 +1,6 @@
 import axios from "axios";
 
-import { buildApiUrl, isSystemProxyBaseUrl, resolveBackendApiUrl, resolveModelRequestConfig, type AiConfig, type ModelChannel } from "@/stores/use-config-store";
+import { buildApiUrl, isSystemProxyBaseUrl, resolveBackendApiUrl, resolveModelRequestConfig, supportsTransparentImageBackground, type AiConfig, type ModelChannel } from "@/stores/use-config-store";
 import { nanoid } from "nanoid";
 import { dataUrlToFile } from "@/lib/image-utils";
 import { buildImageReferencePromptText } from "@/lib/image-reference-prompt";
@@ -233,6 +233,13 @@ function readStatusError(status: number | undefined, fallback: string) {
     return status ? `${fallback}：${status}` : fallback;
 }
 
+function assertTransparentBackgroundSupported(config: AiConfig, model: string) {
+    if (config.transparentBackground !== "true") return;
+    if (!supportsTransparentImageBackground(config, model)) {
+        throw new Error("当前模型不支持透明背景，请关闭“透明背景”或切换至 GPT Image 1 系列模型");
+    }
+}
+
 function withSystemPrompt(config: AiConfig, prompt: string) {
     const systemPrompt = config.systemPrompt.trim();
     return systemPrompt ? `${systemPrompt}\n\n${prompt}` : prompt;
@@ -756,7 +763,9 @@ function parseGeminiImagePayload(payload: GeminiPayload) {
 }
 
 export async function requestGeneration(config: AiConfig, prompt: string, options?: RequestOptions) {
-    const requestConfig = resolveModelRequestConfig(config, config.model || config.imageModel);
+    const selectedModel = config.model || config.imageModel;
+    assertTransparentBackgroundSupported(config, selectedModel);
+    const requestConfig = resolveModelRequestConfig(config, selectedModel);
     const n = Math.max(1, Math.min(15, Math.floor(Math.abs(Number(config.count)) || 1)));
     if (requestConfig.apiFormat === "gemini") {
         try {
@@ -778,6 +787,7 @@ export async function requestGeneration(config: AiConfig, prompt: string, option
                 ...(requestSize ? { size: requestSize } : {}),
                 response_format: "b64_json",
                 output_format: IMAGE_OUTPUT_FORMAT,
+                ...(config.transparentBackground === "true" ? { background: "transparent" } : {}),
             },
             {
                 headers: aiHeaders(requestConfig, "application/json"),
@@ -792,7 +802,9 @@ export async function requestGeneration(config: AiConfig, prompt: string, option
 }
 
 export async function requestEdit(config: AiConfig, prompt: string, references: ReferenceImage[], mask?: ReferenceImage, options?: RequestOptions) {
-    const requestConfig = resolveModelRequestConfig(config, config.model || config.imageModel);
+    const selectedModel = config.model || config.imageModel;
+    assertTransparentBackgroundSupported(config, selectedModel);
+    const requestConfig = resolveModelRequestConfig(config, selectedModel);
     const n = Math.max(1, Math.min(15, Math.floor(Math.abs(Number(config.count)) || 1)));
     const requestPrompt = buildImageReferencePromptText(prompt, references);
     if (requestConfig.apiFormat === "gemini") {
@@ -811,6 +823,9 @@ export async function requestEdit(config: AiConfig, prompt: string, references:
     formData.set("n", String(n));
     formData.set("response_format", "b64_json");
     formData.set("output_format", IMAGE_OUTPUT_FORMAT);
+    if (config.transparentBackground === "true") {
+        formData.set("background", "transparent");
+    }
     if (quality) {
         formData.set("quality", quality);
     }
diff --git a/web/src/stores/use-config-store.ts b/web/src/stores/use-config-store.ts
index 45c346cf4..28de5555b 100644
--- a/web/src/stores/use-config-store.ts
+++ b/web/src/stores/use-config-store.ts
@@ -55,6 +55,7 @@ export type AiConfig = {
     audioModels: string[];
     quality: string;
     size: string;
+    transparentBackground: string;
     count: string;
     canvasImageCount: string;
 };
@@ -101,6 +102,7 @@ export const defaultConfig: AiConfig = {
     audioModels: ["default::gpt-4o-mini-tts"],
     quality: "auto",
     size: "1:1",
+    transparentBackground: "false",
     count: "1",
     canvasImageCount: "1",
 };
@@ -274,6 +276,7 @@ export function normalizeConfigSnapshot(snapshot: ConfigStoreSnapshot) {
             vquality: normalizeVideoResolution(config.vquality),
             videoGenerateAudio: config.videoGenerateAudio || "true",
             videoWatermark: config.videoWatermark || "false",
+            transparentBackground: config.transparentBackground === "true" ? "true" : "false",
             canvasImageCount: config.canvasImageCount || defaultConfig.canvasImageCount,
             imageModels,
             videoModels,
@@ -379,6 +382,13 @@ export function resolveModelChannel(config: AiConfig, value: string) {
     return matched || config.channels[0] || createModelChannel({ id: "default", name: "默认渠道", baseUrl: config.baseUrl, apiKey: config.apiKey, apiFormat: config.apiFormat, models: config.models.map(modelOptionName) });
 }
 
+// 透明背景必须按模型能力显式开放，避免代理接口静默忽略参数后返回不透明图片。
+export function supportsTransparentImageBackground(config: AiConfig, value: string) {
+    const channel = resolveModelChannel(config, value);
+    const model = modelOptionName(value).trim().toLowerCase();
+    return channel.apiFormat === "openai" && (model === "gpt-image-1" || model.startsWith("gpt-image-1-") || model.startsWith("gpt-image-1."));
+}
+
 export function resolveModelRequestConfig(config: AiConfig, value: string) {
     const channel = resolveModelChannel(config, value);
     return {
diff --git a/web/src/types/canvas.ts b/web/src/types/canvas.ts
index 61c516eda..8f44813ed 100644
--- a/web/src/types/canvas.ts
+++ b/web/src/types/canvas.ts
@@ -146,6 +146,7 @@ export type CanvasNodeMetadata = {
     model?: string;
     size?: string;
     quality?: string;
+    transparentBackground?: string;
     count?: number;
     seconds?: string;
     vquality?: string;