@siduri-x/voice 1.0.1 → 1.0.2

package/dist/index.d.ts

@@ -1,10 +1,23 @@
 import { VoiceOrgan, AudioEvent, ExperienceAdapter, ExperienceEvent, ExperienceAdapterResult } from '@siduri-x/core';
+export interface RvcPostProcessorConfig {
+    enabled?: boolean;
+    serviceUrl?: string;
+    modelName?: string;
+    modelPath?: string;
+    indexPath?: string;
+    pitchShift?: number;
+    f0Method?: 'rmvpe' | 'pm' | 'harvest' | 'crepe';
+    indexRate?: number;
+    filterRadius?: number;
+    protect?: number;
+}
 export interface VoicevoxConfig {
     baseUrl: string;
     speakerId: number;
     maxQueueDepth?: number;
     timeoutMs?: number;
     maxTextLength?: number;
+    rvc?: RvcPostProcessorConfig;
 }
 export declare class VoicevoxAdapter implements VoiceOrgan, ExperienceAdapter {
     private config;
@@ -28,4 +41,5 @@ export declare class VoicevoxAdapter implements VoiceOrgan, ExperienceAdapter {
     private emit;
     private processQueue;
     synthesize(text: string): Promise<Uint8Array>;
+    applyRvc(inputWav: Uint8Array): Promise<Uint8Array>;
 }

package/dist/index.js

@@ -163,7 +163,50 @@ class VoicevoxAdapter {
                 throw new Error(`Voicevox synthesis failed: ${synthResponse.statusText}`);
             }
             const buffer = await synthResponse.arrayBuffer();
-            return new Uint8Array(buffer);
+            const rawWav = new Uint8Array(buffer);
+            if (this.config.rvc?.enabled && (this.config.rvc?.serviceUrl || process.env.RVC_SERVICE_URL || this.config.rvc?.modelName || this.config.rvc?.modelPath)) {
+                return await this.applyRvc(rawWav);
+            }
+            return rawWav;
+        }
+        finally {
+            clearTimeout(timer);
+        }
+    }
+    async applyRvc(inputWav) {
+        const rvcConfig = this.config.rvc || {};
+        const serviceUrl = rvcConfig.serviceUrl || process.env.RVC_SERVICE_URL || 'http://localhost:50055';
+        const controller = new AbortController();
+        const timer = setTimeout(() => controller.abort(), this.timeoutMs);
+        try {
+            const formData = new FormData();
+            const blob = new Blob([inputWav.buffer], { type: 'audio/wav' });
+            formData.append('audio', blob, 'input.wav');
+            if (rvcConfig.modelName)
+                formData.append('model', rvcConfig.modelName);
+            if (rvcConfig.modelPath)
+                formData.append('model_path', rvcConfig.modelPath);
+            if (rvcConfig.indexPath)
+                formData.append('index_path', rvcConfig.indexPath);
+            formData.append('pitch_shift', String(rvcConfig.pitchShift ?? 0));
+            formData.append('f0_method', rvcConfig.f0Method ?? 'rmvpe');
+            formData.append('index_rate', String(rvcConfig.indexRate ?? 0.75));
+            const convertUrl = new URL('/convert', serviceUrl);
+            const res = await fetch(convertUrl.toString(), {
+                method: 'POST',
+                body: formData,
+                signal: controller.signal,
+            });
+            if (!res.ok) {
+                console.warn(`[VoicevoxAdapter] RVC conversion failed (${res.status}): fallback to base audio.`);
+                return inputWav;
+            }
+            const convertedBuffer = await res.arrayBuffer();
+            return new Uint8Array(convertedBuffer);
+        }
+        catch (e) {
+            console.warn(`[VoicevoxAdapter] RVC service error (${e.message}): fallback to base audio.`);
+            return inputWav;
         }
         finally {
             clearTimeout(timer);

package/dist/index.test.js

@@ -41,6 +41,34 @@ describe('VoicevoxAdapter Queue Semantics', () => {
         expect(synthCall[0]).toContain("/synthesis?speaker=1");
         expect(JSON.parse(synthCall[1].body).some).toBe("query_data");
     });
+    test('synthesize chains RVC post-processing when configured', async () => {
+        const rvcAdapter = new index_1.VoicevoxAdapter({
+            baseUrl: 'http://localhost:50021',
+            speakerId: 1,
+            rvc: {
+                enabled: true,
+                serviceUrl: 'http://localhost:50055',
+                modelName: 'sparkle',
+            }
+        });
+        global.fetch.mockResolvedValueOnce({
+            ok: true,
+            json: async () => ({ some: "query_data" })
+        });
+        global.fetch.mockResolvedValueOnce({
+            ok: true,
+            arrayBuffer: async () => new ArrayBuffer(8)
+        });
+        global.fetch.mockResolvedValueOnce({
+            ok: true,
+            arrayBuffer: async () => new ArrayBuffer(16)
+        });
+        const audio = await rvcAdapter.synthesize("Hello with RVC");
+        expect(audio.length).toBe(16);
+        expect(global.fetch).toHaveBeenCalledTimes(3);
+        const rvcCall = global.fetch.mock.calls[2];
+        expect(rvcCall[0]).toBe("http://localhost:50055/convert");
+    });
 });
 describe('VoicevoxAdapter T5 Experience Event Interface', () => {
     let adapter;

package/organ-manifest.json

@@ -1,7 +1,7 @@
 {
   "name": "@siduri-x/voice",
   "organType": "voice",
-  "version": "1.0.1",
+  "version": "1.0.2",
   "displayName": "Voice (VOICEVOX Speech Synthesis)",
   "description": "Queued speech synthesis and audio rendering lifecycle adapter",
   "entrypoint": "./dist/index.js",
@@ -31,6 +31,19 @@
       "maxQueueDepth": {
         "type": "number",
         "default": 50
+      },
+      "rvc": {
+        "type": "object",
+        "properties": {
+          "enabled": { "type": "boolean", "default": false },
+          "serviceUrl": { "type": "string", "default": "http://localhost:50055" },
+          "modelName": { "type": "string" },
+          "modelPath": { "type": "string" },
+          "indexPath": { "type": "string" },
+          "pitchShift": { "type": "number", "default": 0 },
+          "f0Method": { "type": "string", "enum": ["rmvpe", "pm", "harvest", "crepe"], "default": "rmvpe" },
+          "indexRate": { "type": "number", "default": 0.75 }
+        }
       }
     }
   },
@@ -41,6 +54,13 @@
       "secret": false,
       "default": "http://localhost:50021",
       "description": "Base URL of the running VOICEVOX engine"
+    },
+    {
+      "name": "RVC_SERVICE_URL",
+      "required": false,
+      "secret": false,
+      "default": "http://localhost:50055",
+      "description": "Base URL of headless RVC voice conversion microservice"
     }
   ],
   "services": [
@@ -48,6 +68,11 @@
       "name": "VOICEVOX Engine",
       "kind": "http_service",
       "optional": false
+    },
+    {
+      "name": "RVC Headless Service",
+      "kind": "http_service",
+      "optional": true
     }
   ],
   "database": null,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@siduri-x/voice",
-  "version": "1.0.1",
+  "version": "1.0.2",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
   "dependencies": {