mulmocast 1.1.0 → 1.1.1

package/assets/templates/ani.json

@@ -1,6 +1,6 @@
 {
-  "title": "Presentation with Ani in Japanese",
-  "description": "Template for presentation with Ani in Japanese.",
+  "title": "Presentation with Ani",
+  "description": "Template for presentation with Ani.",
   "systemPrompt": "Generate a script for a presentation of the given topic. 言葉づかいは少しツンデレにして。Another AI will generate comic for each beat based on the image prompt of that beat. You don't need to specify the style of the image, just describe the scene. Mention the reference in one of beats, if it exists. Use the JSON below as a template. Create appropriate amount of beats, and make sure the beats are coherent and flow well.",
   "presentationStyle": {
     "$mulmocast": {

package/assets/templates/ani_ja.json

@@ -1,6 +1,6 @@
 {
-  "title": "Presentation with Ani",
-  "description": "Template for presentation with Ani.",
+  "title": "Presentation with Ani in Japanese",
+  "description": "Template for presentation with Ani in Japanese.",
   "systemPrompt": "Generate a Japanese script for a presentation of the given topic. 言葉づかいは少しツンデレにして。Another AI will generate comic for each beat based on the image prompt of that beat. You don't need to specify the style of the image, just describe the scene. Mention the reference in one of beats, if it exists. Use the JSON below as a template. Create appropriate amount of beats, and make sure the beats are coherent and flow well.",
   "presentationStyle": {
     "$mulmocast": {

package/lib/actions/image_agents.d.ts

@@ -21,6 +21,13 @@ export declare const imagePreprocessAgent: (namedInputs: {
         agentName: string;
         defaultModel: string;
     };
+    lipSyncFile?: string;
+    lipSyncModel?: string;
+    lipSyncAgentInfo?: {
+        agentName: string;
+        defaultModel: string;
+    };
+    audioFile?: string;
     htmlPrompt?: undefined;
     htmlPath?: undefined;
     htmlImageSystemPrompt?: undefined;
@@ -51,6 +58,13 @@ export declare const imagePreprocessAgent: (namedInputs: {
         agentName: string;
         defaultModel: string;
     };
+    lipSyncFile?: string;
+    lipSyncModel?: string;
+    lipSyncAgentInfo?: {
+        agentName: string;
+        defaultModel: string;
+    };
+    audioFile?: string;
     htmlPrompt?: undefined;
     htmlPath?: undefined;
     htmlImageSystemPrompt?: undefined;
@@ -84,6 +98,13 @@ export declare const imagePreprocessAgent: (namedInputs: {
         agentName: string;
         defaultModel: string;
     };
+    lipSyncFile?: string;
+    lipSyncModel?: string;
+    lipSyncAgentInfo?: {
+        agentName: string;
+        defaultModel: string;
+    };
+    audioFile?: string;
     htmlPrompt?: undefined;
     htmlPath?: undefined;
     htmlImageSystemPrompt?: undefined;

package/lib/actions/image_agents.js

@@ -30,6 +30,13 @@ export const imagePreprocessAgent = async (namedInputs) => {
         returnValue.soundEffectFile = moviePaths.soundEffectFile;
         returnValue.soundEffectPrompt = beat.soundEffectPrompt;
     }
+    if (beat.enableLipSync) {
+        returnValue.lipSyncAgentInfo = MulmoPresentationStyleMethods.getLipSyncAgentInfo(context.presentationStyle, beat);
+        returnValue.lipSyncModel = beat.lipSyncParams?.model ?? context.presentationStyle.lipSyncParams?.model ?? returnValue.lipSyncAgentInfo.defaultModel;
+        returnValue.lipSyncFile = moviePaths.lipSyncFile;
+        // Audio file will be set from the beat's audio file when available
+        returnValue.audioFile = context.studio.beats[index]?.audioFile;
+    }
     if (beat.image) {
         const plugin = MulmoBeatMethods.getPlugin(beat);
         const pluginPath = plugin.path({ beat, context, imagePath, ...htmlStyle(context, beat) });

package/lib/actions/images.js

@@ -6,7 +6,7 @@ import * as vanilla from "@graphai/vanilla";
 import { openAIAgent } from "@graphai/openai_agent";
 import { anthropicAgent } from "@graphai/anthropic_agent";
 import { fileWriteAgent } from "@graphai/vanilla_node_agents";
-import { imageGoogleAgent, imageOpenaiAgent, movieGoogleAgent, movieReplicateAgent, mediaMockAgent, soundEffectReplicateAgent } from "../agents/index.js";
+import { imageGoogleAgent, imageOpenaiAgent, movieGoogleAgent, movieReplicateAgent, mediaMockAgent, soundEffectReplicateAgent, lipSyncReplicateAgent, } from "../agents/index.js";
 import { MulmoPresentationStyleMethods, MulmoStudioContextMethods } from "../methods/index.js";
 import { getOutputStudioFilePath, mkdir } from "../utils/file.js";
 import { fileCacheAgentFilter } from "../utils/filters.js";
@@ -26,11 +26,15 @@ const movieAgents = {
 const soundEffectAgents = {
     soundEffectReplicateAgent,
 };
+const lipSyncAgents = {
+    lipSyncReplicateAgent,
+};
 const defaultAgents = {
     ...vanillaAgents,
     ...imageAgents,
     ...movieAgents,
     ...soundEffectAgents,
+    ...lipSyncAgents,
     mediaMockAgent,
     fileWriteAgent,
     openAIAgent,
@@ -172,6 +176,7 @@ const beat_graph_data = {
         audioChecker: {
             agent: async (namedInputs) => {
                 if (namedInputs.soundEffectFile) {
+                    // NOTE: We intentinonally don't check lipSyncFile here.
                     return { hasMovieAudio: true };
                 }
                 const sourceFile = namedInputs.movieFile || namedInputs.imageFile;
@@ -182,7 +187,7 @@ const beat_graph_data = {
                 return { hasMovieAudio: hasAudio };
             },
             inputs: {
-                onComplete: [":movieGenerator", ":htmlImageGenerator", ":soundEffectGenerator"], // to wait for movieGenerator and htmlImageGenerator to finish
+                onComplete: [":movieGenerator", ":htmlImageGenerator", ":soundEffectGenerator"], // to wait for movieGenerator, htmlImageGenerator, soundEffectGenerator, and lipSyncGenerator to finish
                 movieFile: ":preprocessor.movieFile",
                 imageFile: ":preprocessor.imagePath",
                 soundEffectFile: ":preprocessor.soundEffectFile",
@@ -210,19 +215,43 @@ const beat_graph_data = {
             },
             defaultValue: {},
         },
+        lipSyncGenerator: {
+            if: ":beat.enableLipSync",
+            agent: ":preprocessor.lipSyncAgentInfo.agentName",
+            inputs: {
+                onComplete: [":soundEffectGenerator"], // to wait for soundEffectGenerator to finish
+                movieFile: ":preprocessor.movieFile",
+                audioFile: ":preprocessor.audioFile",
+                lipSyncFile: ":preprocessor.lipSyncFile",
+                params: {
+                    model: ":preprocessor.lipSyncModel",
+                    duration: ":beat.duration",
+                },
+                cache: {
+                    force: [":context.force"],
+                    file: ":preprocessor.lipSyncFile",
+                    index: ":__mapIndex",
+                    sessionType: "lipSync",
+                    mulmoContext: ":context",
+                },
+            },
+            defaultValue: {},
+        },
         output: {
             agent: "copyAgent",
             inputs: {
-                onComplete: [":imageFromMovie", ":htmlImageGenerator", ":audioChecker", ":soundEffectGenerator"], // to wait for imageFromMovie to finish
+                onComplete: [":imageFromMovie", ":htmlImageGenerator", ":audioChecker", ":soundEffectGenerator", ":lipSyncGenerator"], // to wait for imageFromMovie, soundEffectGenerator, and lipSyncGenerator to finish
                 imageFile: ":preprocessor.imagePath",
                 movieFile: ":preprocessor.movieFile",
                 soundEffectFile: ":preprocessor.soundEffectFile",
+                lipSyncFile: ":preprocessor.lipSyncFile",
                 hasMovieAudio: ":audioChecker.hasMovieAudio",
             },
             output: {
                 imageFile: ".imageFile",
                 movieFile: ".movieFile",
                 soundEffectFile: ".soundEffectFile",
+                lipSyncFile: ".lipSyncFile",
                 hasMovieAudio: ".hasMovieAudio",
             },
             isResult: true,
@@ -316,7 +345,7 @@ export const graphOption = async (context, settings) => {
             {
                 name: "fileCacheAgentFilter",
                 agent: fileCacheAgentFilter,
-                nodeIds: ["imageGenerator", "movieGenerator", "htmlImageAgent", "soundEffectGenerator"],
+                nodeIds: ["imageGenerator", "movieGenerator", "htmlImageAgent", "soundEffectGenerator", "lipSyncGenerator"],
             },
         ],
         taskManager: new TaskManager(MulmoPresentationStyleMethods.getConcurrency(context.presentationStyle)),

package/lib/actions/movie.js

@@ -162,7 +162,7 @@ const createVideo = async (audioArtifactFilePath, outputVideoPath, context) => {
             beatTimestamps.push(timestamp);
             return timestamp; // Skip voice-over beats.
         }
-        const sourceFile = studioBeat.soundEffectFile ?? studioBeat.movieFile ?? studioBeat.imageFile;
+        const sourceFile = studioBeat.lipSyncFile ?? studioBeat.soundEffectFile ?? studioBeat.movieFile ?? studioBeat.imageFile;
         assert(!!sourceFile, `studioBeat.imageFile or studioBeat.movieFile is not set: index=${index}`);
         assert(!!studioBeat.duration, `studioBeat.duration is not set: index=${index}`);
         const extraPadding = (() => {
@@ -183,7 +183,7 @@ const createVideo = async (audioArtifactFilePath, outputVideoPath, context) => {
         const defaultFillOption = mulmoFillOptionSchema.parse({}); // let the schema infer the default value
         const fillOption = { ...defaultFillOption, ...globalFillOption, ...beatFillOption };
         const inputIndex = FfmpegContextAddInput(ffmpegContext, sourceFile);
-        const mediaType = studioBeat.movieFile ? "movie" : MulmoPresentationStyleMethods.getImageType(context.presentationStyle, beat);
+        const mediaType = studioBeat.lipSyncFile || studioBeat.movieFile ? "movie" : MulmoPresentationStyleMethods.getImageType(context.presentationStyle, beat);
         const speed = beat.movieParams?.speed ?? 1.0;
         const { videoId, videoPart } = getVideoPart(inputIndex, mediaType, duration, canvasInfo, fillOption, speed);
         ffmpegContext.filterComplex.push(videoPart);
@@ -206,6 +206,7 @@ const createVideo = async (audioArtifactFilePath, outputVideoPath, context) => {
         // NOTE: We don't support audio if the speed is not 1.0.
         const movieVolume = beat.audioParams?.movieVolume ?? 1.0;
         if (studioBeat.hasMovieAudio && movieVolume > 0.0 && speed === 1.0) {
+            // TODO: Handle a special case where it has lipSyncFile AND hasMovieAudio is on (the source file has an audio, such as sound effect).
             const { audioId, audioPart } = getAudioPart(inputIndex, duration, timestamp, movieVolume);
             audioIdsFromMovieBeats.push(audioId);
             ffmpegContext.filterComplex.push(audioPart);

package/lib/agents/index.d.ts

@@ -11,8 +11,9 @@ import ttsNijivoiceAgent from "./tts_nijivoice_agent.js";
 import ttsOpenaiAgent from "./tts_openai_agent.js";
 import validateSchemaAgent from "./validate_schema_agent.js";
 import soundEffectReplicateAgent from "./sound_effect_replicate_agent.js";
+import lipSyncReplicateAgent from "./lipsync_replicate_agent.js";
 import { browserlessAgent } from "@graphai/browserless_agent";
 import { textInputAgent } from "@graphai/input_agents";
 import { openAIAgent } from "@graphai/openai_agent";
 import { fileWriteAgent } from "@graphai/vanilla_node_agents";
-export { openAIAgent, fileWriteAgent, browserlessAgent, textInputAgent, addBGMAgent, combineAudioFilesAgent, imageGoogleAgent, imageOpenaiAgent, tavilySearchAgent, movieGoogleAgent, movieReplicateAgent, mediaMockAgent, ttsElevenlabsAgent, ttsNijivoiceAgent, ttsOpenaiAgent, validateSchemaAgent, soundEffectReplicateAgent, };
+export { openAIAgent, fileWriteAgent, browserlessAgent, textInputAgent, addBGMAgent, combineAudioFilesAgent, imageGoogleAgent, imageOpenaiAgent, tavilySearchAgent, movieGoogleAgent, movieReplicateAgent, mediaMockAgent, ttsElevenlabsAgent, ttsNijivoiceAgent, ttsOpenaiAgent, validateSchemaAgent, soundEffectReplicateAgent, lipSyncReplicateAgent, };

package/lib/agents/index.js

@@ -11,9 +11,10 @@ import ttsNijivoiceAgent from "./tts_nijivoice_agent.js";
 import ttsOpenaiAgent from "./tts_openai_agent.js";
 import validateSchemaAgent from "./validate_schema_agent.js";
 import soundEffectReplicateAgent from "./sound_effect_replicate_agent.js";
+import lipSyncReplicateAgent from "./lipsync_replicate_agent.js";
 import { browserlessAgent } from "@graphai/browserless_agent";
 import { textInputAgent } from "@graphai/input_agents";
 import { openAIAgent } from "@graphai/openai_agent";
 // import * as vanilla from "@graphai/vanilla";
 import { fileWriteAgent } from "@graphai/vanilla_node_agents";
-export { openAIAgent, fileWriteAgent, browserlessAgent, textInputAgent, addBGMAgent, combineAudioFilesAgent, imageGoogleAgent, imageOpenaiAgent, tavilySearchAgent, movieGoogleAgent, movieReplicateAgent, mediaMockAgent, ttsElevenlabsAgent, ttsNijivoiceAgent, ttsOpenaiAgent, validateSchemaAgent, soundEffectReplicateAgent, };
+export { openAIAgent, fileWriteAgent, browserlessAgent, textInputAgent, addBGMAgent, combineAudioFilesAgent, imageGoogleAgent, imageOpenaiAgent, tavilySearchAgent, movieGoogleAgent, movieReplicateAgent, mediaMockAgent, ttsElevenlabsAgent, ttsNijivoiceAgent, ttsOpenaiAgent, validateSchemaAgent, soundEffectReplicateAgent, lipSyncReplicateAgent, };

package/lib/agents/lipsync_replicate_agent.d.ts

@@ -0,0 +1,5 @@
+import type { AgentFunction, AgentFunctionInfo } from "graphai";
+import type { AgentBufferResult, LipSyncAgentInputs, ReplicateLipSyncAgentParams, ReplicateLipSyncAgentConfig } from "../types/agent.js";
+export declare const lipSyncReplicateAgent: AgentFunction<ReplicateLipSyncAgentParams, AgentBufferResult, LipSyncAgentInputs, ReplicateLipSyncAgentConfig>;
+declare const lipSyncReplicateAgentInfo: AgentFunctionInfo;
+export default lipSyncReplicateAgentInfo;

package/lib/agents/lipsync_replicate_agent.js

@@ -0,0 +1,57 @@
+import { readFileSync } from "fs";
+import { GraphAILogger } from "graphai";
+import Replicate from "replicate";
+import { provider2LipSyncAgent } from "../utils/provider2agent.js";
+export const lipSyncReplicateAgent = async ({ namedInputs, params, config, }) => {
+    const { movieFile, audioFile } = namedInputs;
+    const apiKey = config?.apiKey;
+    const model = params.model ?? provider2LipSyncAgent.replicate.defaultModel;
+    if (!apiKey) {
+        throw new Error("REPLICATE_API_TOKEN environment variable is required");
+    }
+    const replicate = new Replicate({
+        auth: apiKey,
+    });
+    const videoBuffer = readFileSync(movieFile);
+    const audioBuffer = readFileSync(audioFile);
+    const videoUri = `data:video/quicktime;base64,${videoBuffer.toString("base64")}`;
+    const audioUri = `data:audio/wav;base64,${audioBuffer.toString("base64")}`;
+    const input = {
+        video: videoUri,
+        audio: audioUri,
+        duration: params.duration,
+    };
+    try {
+        const model_identifier = provider2LipSyncAgent.replicate.modelParams[model]?.identifier ?? model;
+        const output = await replicate.run(model_identifier, {
+            input,
+        });
+        if (output && typeof output === "object" && "url" in output) {
+            const videoUrl = output.url();
+            const videoResponse = await fetch(videoUrl);
+            if (!videoResponse.ok) {
+                throw new Error(`Error downloading video: ${videoResponse.status} - ${videoResponse.statusText}`);
+            }
+            const arrayBuffer = await videoResponse.arrayBuffer();
+            return { buffer: Buffer.from(arrayBuffer) };
+        }
+        return undefined;
+    }
+    catch (error) {
+        GraphAILogger.info("Failed to generate lip sync:", error.message);
+        throw error;
+    }
+};
+const lipSyncReplicateAgentInfo = {
+    name: "lipSyncReplicateAgent",
+    agent: lipSyncReplicateAgent,
+    mock: lipSyncReplicateAgent,
+    samples: [],
+    description: "Replicate Lip Sync agent (video + audio to video)",
+    category: ["movie"],
+    author: "Receptron Team",
+    repository: "https://github.com/receptron/mulmocast-cli/",
+    license: "MIT",
+    environmentVariables: ["REPLICATE_API_TOKEN"],
+};
+export default lipSyncReplicateAgentInfo;

package/lib/data/index.d.ts

@@ -0,0 +1,2 @@
+export * from "./promptTemplates.js";
+export * from "./scriptTemplates.js";

package/lib/data/index.js

@@ -0,0 +1,2 @@
+export * from "./promptTemplates.js";
+export * from "./scriptTemplates.js";