mulmocast 1.1.5 → 1.1.7

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (77) hide show
  1. package/lib/actions/audio.js +10 -1
  2. package/lib/actions/image_agents.d.ts +3 -12
  3. package/lib/actions/image_agents.js +12 -8
  4. package/lib/actions/images.js +2 -1
  5. package/lib/actions/translate.d.ts +51 -2
  6. package/lib/actions/translate.js +193 -148
  7. package/lib/agents/combine_audio_files_agent.js +1 -1
  8. package/lib/agents/lipsync_replicate_agent.js +10 -3
  9. package/lib/agents/tts_nijivoice_agent.js +1 -1
  10. package/lib/cli/commands/audio/handler.js +1 -1
  11. package/lib/cli/commands/image/handler.js +1 -1
  12. package/lib/cli/commands/movie/handler.js +1 -1
  13. package/lib/cli/commands/pdf/handler.js +1 -1
  14. package/lib/cli/helpers.d.ts +1 -4
  15. package/lib/cli/helpers.js +3 -2
  16. package/lib/index.common.d.ts +1 -0
  17. package/lib/index.common.js +1 -0
  18. package/lib/mcp/server.js +1 -1
  19. package/lib/methods/mulmo_presentation_style.d.ts +3 -2
  20. package/lib/methods/mulmo_script.d.ts +4 -1
  21. package/lib/methods/mulmo_script.js +18 -2
  22. package/lib/methods/mulmo_studio_context.d.ts +1 -0
  23. package/lib/methods/mulmo_studio_context.js +8 -0
  24. package/lib/types/agent.d.ts +1 -0
  25. package/lib/types/schema.d.ts +326 -230
  26. package/lib/types/schema.js +10 -3
  27. package/lib/types/type.d.ts +3 -2
  28. package/lib/utils/const.d.ts +1 -0
  29. package/lib/utils/const.js +2 -1
  30. package/lib/utils/context.d.ts +393 -50
  31. package/lib/utils/context.js +90 -57
  32. package/lib/utils/filters.d.ts +1 -0
  33. package/lib/utils/filters.js +8 -0
  34. package/lib/utils/image_plugins/mermaid.js +1 -1
  35. package/lib/utils/image_plugins/source.js +1 -1
  36. package/lib/utils/preprocess.d.ts +2 -2
  37. package/lib/utils/preprocess.js +3 -3
  38. package/lib/utils/provider2agent.d.ts +3 -2
  39. package/lib/utils/provider2agent.js +20 -2
  40. package/lib/utils/string.d.ts +1 -1
  41. package/lib/utils/string.js +12 -8
  42. package/lib/utils/utils.js +2 -6
  43. package/package.json +2 -2
  44. package/scripts/templates/image_refs.json +1 -0
  45. package/scripts/templates/voice_over.json +1 -0
  46. package/scripts/test/gpt.json +1 -0
  47. package/scripts/test/test1.json +1 -0
  48. package/scripts/test/test_audio.json +1 -0
  49. package/scripts/test/test_audio_instructions.json +1 -0
  50. package/scripts/test/test_beats.json +1 -0
  51. package/scripts/test/test_captions.json +1 -0
  52. package/scripts/test/test_elevenlabs_models.json +1 -0
  53. package/scripts/test/test_hello.json +1 -0
  54. package/scripts/test/test_hello_google.json +1 -0
  55. package/scripts/test/test_html.json +1 -0
  56. package/scripts/test/test_image_refs.json +1 -0
  57. package/scripts/test/test_images.json +1 -0
  58. package/scripts/test/test_lang.json +58 -2
  59. package/scripts/test/test_layout.json +1 -0
  60. package/scripts/test/test_lipsync.json +9 -0
  61. package/scripts/test/test_loop.json +1 -0
  62. package/scripts/test/test_media.json +1 -0
  63. package/scripts/test/test_mixed_providers.json +1 -0
  64. package/scripts/test/test_movie.json +1 -0
  65. package/scripts/test/test_no_audio.json +1 -0
  66. package/scripts/test/test_no_audio_with_credit.json +1 -0
  67. package/scripts/test/test_order.json +1 -0
  68. package/scripts/test/test_order_portrait.json +1 -0
  69. package/scripts/test/test_replicate.json +19 -0
  70. package/scripts/test/test_slideout_left_no_audio.json +1 -0
  71. package/scripts/test/test_spillover.json +1 -0
  72. package/scripts/test/test_transition.json +1 -0
  73. package/scripts/test/test_transition_no_audio.json +1 -0
  74. package/scripts/test/test_video_speed.json +1 -0
  75. package/scripts/test/test_voice_over.json +1 -0
  76. package/scripts/test/test_voices.json +1 -0
  77. package/scripts/templates/image_prompt_only_template.ts +0 -95
@@ -1,44 +1,27 @@
1
1
  import { GraphAILogger } from "graphai";
2
2
  import fs from "fs";
3
3
  import { readMulmoScriptFile, fetchMulmoScriptFile } from "./file.js";
4
- import { createOrUpdateStudioData } from "./preprocess.js";
5
- import { mulmoPresentationStyleSchema, mulmoStudioMultiLingualSchema } from "../types/index.js";
6
- export const fetchScript = async (isHttpPath, mulmoFilePath, fileOrUrl) => {
7
- if (isHttpPath) {
8
- const res = await fetchMulmoScriptFile(fileOrUrl);
9
- if (!res.result || !res.script) {
10
- GraphAILogger.info(`ERROR: HTTP error! ${res.status} ${fileOrUrl}`);
11
- return null;
12
- }
13
- return res.script;
14
- }
15
- if (!fs.existsSync(mulmoFilePath)) {
16
- GraphAILogger.info(`ERROR: File not exists ${mulmoFilePath}`);
17
- return null;
18
- }
19
- return readMulmoScriptFile(mulmoFilePath, "ERROR: File does not exist " + mulmoFilePath)?.mulmoData ?? null;
20
- };
21
- export const getMultiLingual = (multilingualFilePath, studioBeatsLength) => {
22
- if (fs.existsSync(multilingualFilePath)) {
23
- const jsonData = readMulmoScriptFile(multilingualFilePath, "ERROR: File does not exist " + multilingualFilePath)?.mulmoData ?? null;
24
- const dataSet = mulmoStudioMultiLingualSchema.parse(jsonData);
25
- while (dataSet.length < studioBeatsLength) {
26
- dataSet.push({ multiLingualTexts: {} });
27
- }
28
- dataSet.length = studioBeatsLength;
29
- return dataSet;
30
- }
31
- return [...Array(studioBeatsLength)].map(() => ({ multiLingualTexts: {} }));
32
- };
33
- export const getPresentationStyle = (presentationStylePath) => {
34
- if (presentationStylePath) {
35
- if (!fs.existsSync(presentationStylePath)) {
36
- throw new Error(`ERROR: File not exists ${presentationStylePath}`);
37
- }
38
- const jsonData = readMulmoScriptFile(presentationStylePath, "ERROR: File does not exist " + presentationStylePath)?.mulmoData ?? null;
39
- return mulmoPresentationStyleSchema.parse(jsonData);
40
- }
41
- return null;
4
+ import { mulmoStudioSchema, mulmoCaptionParamsSchema, mulmoPresentationStyleSchema } from "../types/schema.js";
5
+ import { MulmoPresentationStyleMethods, MulmoScriptMethods, MulmoStudioMultiLingualMethod } from "../methods/index.js";
6
+ const mulmoCredit = (speaker) => {
7
+ return {
8
+ speaker,
9
+ text: "",
10
+ image: {
11
+ type: "image",
12
+ source: {
13
+ kind: "url",
14
+ url: "https://github.com/receptron/mulmocast-cli/raw/refs/heads/main/assets/images/mulmocast_credit.png",
15
+ },
16
+ },
17
+ audio: {
18
+ type: "audio",
19
+ source: {
20
+ kind: "url",
21
+ url: "https://github.com/receptron/mulmocast-cli/raw/refs/heads/main/assets/audio/silent300.mp3",
22
+ },
23
+ },
24
+ };
42
25
  };
43
26
  const initSessionState = () => {
44
27
  return {
@@ -63,32 +46,82 @@ const initSessionState = () => {
63
46
  },
64
47
  };
65
48
  };
66
- const buildContext = (studio, files, presentationStyle, multiLingual, force, lang) => {
67
- return {
68
- studio,
69
- fileDirs: files,
70
- force: Boolean(force),
71
- lang,
72
- sessionState: initSessionState(),
73
- presentationStyle: presentationStyle ?? studio.script,
74
- multiLingual,
75
- };
49
+ export const createStudioData = (_mulmoScript, fileName, videoCaptionLang, presentationStyle) => {
50
+ // validate and insert default value
51
+ const mulmoScript = _mulmoScript.__test_invalid__ ? _mulmoScript : MulmoScriptMethods.validate(_mulmoScript);
52
+ // We need to parse it to fill default values
53
+ const studio = mulmoStudioSchema.parse({
54
+ script: mulmoScript,
55
+ filename: fileName,
56
+ beats: [...Array(mulmoScript.beats.length)].map(() => ({})),
57
+ });
58
+ // TODO: Move this code out of this function later
59
+ // Addition cloing credit
60
+ if (mulmoScript.$mulmocast.credit === "closing") {
61
+ const defaultSpeaker = MulmoPresentationStyleMethods.getDefaultSpeaker(presentationStyle ?? studio.script);
62
+ mulmoScript.beats.push(mulmoCredit(mulmoScript.beats[0].speaker ?? defaultSpeaker)); // First speaker
63
+ }
64
+ studio.script = MulmoScriptMethods.validate(mulmoScript); // update the script
65
+ studio.beats = studio.script.beats.map((_, index) => studio.beats[index] ?? {});
66
+ if (videoCaptionLang) {
67
+ studio.script.captionParams = mulmoCaptionParamsSchema.parse({
68
+ ...(studio.script.captionParams ?? {}),
69
+ lang: videoCaptionLang,
70
+ });
71
+ }
72
+ return studio;
73
+ };
74
+ export const fetchScript = async (isHttpPath, mulmoFilePath, fileOrUrl) => {
75
+ if (isHttpPath) {
76
+ const res = await fetchMulmoScriptFile(fileOrUrl);
77
+ if (!res.result || !res.script) {
78
+ GraphAILogger.info(`ERROR: HTTP error! ${res.status} ${fileOrUrl}`);
79
+ return null;
80
+ }
81
+ return res.script;
82
+ }
83
+ if (!fs.existsSync(mulmoFilePath)) {
84
+ GraphAILogger.info(`ERROR: File not exists ${mulmoFilePath}`);
85
+ return null;
86
+ }
87
+ return readMulmoScriptFile(mulmoFilePath, "ERROR: File does not exist " + mulmoFilePath)?.mulmoData ?? null;
88
+ };
89
+ export const getMultiLingual = (multilingualFilePath, studioBeatsLength) => {
90
+ if (!fs.existsSync(multilingualFilePath)) {
91
+ return [...Array(studioBeatsLength)].map(() => ({ multiLingualTexts: {} }));
92
+ }
93
+ const jsonData = readMulmoScriptFile(multilingualFilePath, "ERROR: File does not exist " + multilingualFilePath)?.mulmoData ?? null;
94
+ return MulmoStudioMultiLingualMethod.validate(jsonData, studioBeatsLength);
95
+ };
96
+ export const getPresentationStyle = (presentationStylePath) => {
97
+ if (!presentationStylePath) {
98
+ return null;
99
+ }
100
+ if (!fs.existsSync(presentationStylePath)) {
101
+ throw new Error(`ERROR: File not exists ${presentationStylePath}`);
102
+ }
103
+ const jsonData = readMulmoScriptFile(presentationStylePath, "ERROR: File does not exist " + presentationStylePath)?.mulmoData ?? null;
104
+ return mulmoPresentationStyleSchema.parse(jsonData);
76
105
  };
77
- export const initializeContextFromFiles = async (files, raiseError, force, caption, lang) => {
78
- const { fileName, isHttpPath, fileOrUrl, mulmoFilePath, outputStudioFilePath, presentationStylePath, outputMultilingualFilePath } = files;
79
- // read mulmoScript, presentationStyle, currentStudio from files
106
+ export const initializeContextFromFiles = async (files, raiseError, force, captionLang, targetLang) => {
107
+ const { fileName, isHttpPath, fileOrUrl, mulmoFilePath, presentationStylePath, outputMultilingualFilePath } = files;
80
108
  const mulmoScript = await fetchScript(isHttpPath, mulmoFilePath, fileOrUrl);
81
109
  if (!mulmoScript) {
82
110
  return null;
83
111
  }
84
- const presentationStyle = getPresentationStyle(presentationStylePath);
85
- // Create or update MulmoStudio file with MulmoScript
86
- const currentStudio = readMulmoScriptFile(outputStudioFilePath);
87
112
  try {
88
- // validate mulmoStudioSchema. skip if __test_invalid__ is true
89
- const studio = createOrUpdateStudioData(mulmoScript, currentStudio?.mulmoData, fileName, caption, presentationStyle);
113
+ const presentationStyle = getPresentationStyle(presentationStylePath);
114
+ const studio = createStudioData(mulmoScript, fileName, captionLang, presentationStyle);
90
115
  const multiLingual = getMultiLingual(outputMultilingualFilePath, studio.beats.length);
91
- return buildContext(studio, files, presentationStyle, multiLingual, force, lang);
116
+ return {
117
+ studio,
118
+ multiLingual,
119
+ fileDirs: files,
120
+ presentationStyle: presentationStyle ?? studio.script,
121
+ sessionState: initSessionState(),
122
+ force: Boolean(force),
123
+ lang: targetLang ?? studio.script.lang, // This lang is target Language. studio.lang is default Language
124
+ };
92
125
  }
93
126
  catch (error) {
94
127
  GraphAILogger.info(`Error: invalid MulmoScript Schema: ${isHttpPath ? fileOrUrl : mulmoFilePath} \n ${error}`);
@@ -1,4 +1,5 @@
1
1
  import "dotenv/config";
2
2
  import type { AgentFilterFunction } from "graphai";
3
+ export declare const nijovoiceTextAgentFilter: AgentFilterFunction;
3
4
  export declare const fileCacheAgentFilter: AgentFilterFunction;
4
5
  export declare const browserlessCacheGenerator: (cacheDir: string) => AgentFilterFunction;
@@ -6,6 +6,14 @@ import { GraphAILogger } from "graphai";
6
6
  import { writingMessage } from "./file.js";
7
7
  import { text2hash } from "./utils.js";
8
8
  import { MulmoStudioContextMethods } from "../methods/mulmo_studio_context.js";
9
+ import { replacementsJa, replacePairsJa } from "../utils/string.js";
10
+ export const nijovoiceTextAgentFilter = async (context, next) => {
11
+ const { text, provider, lang } = context.namedInputs;
12
+ if (provider === "nijivoice" && lang === "ja") {
13
+ context.namedInputs.text = replacePairsJa(replacementsJa)(text);
14
+ }
15
+ return next(context);
16
+ };
9
17
  export const fileCacheAgentFilter = async (context, next) => {
10
18
  const { force, file, index, mulmoContext, sessionType } = context.namedInputs.cache;
11
19
  const shouldUseCache = async () => {
@@ -5,7 +5,7 @@ import { parrotingImagePath } from "./utils.js";
5
5
  export const imageType = "mermaid";
6
6
  const processMermaid = async (params) => {
7
7
  const { beat, imagePath, canvasSize, context, textSlideStyle } = params;
8
- if (!beat || !beat.image || beat.image.type !== imageType)
8
+ if (!beat?.image || beat.image.type !== imageType)
9
9
  return;
10
10
  const template = getHTMLFile("mermaid");
11
11
  const diagram_code = await MulmoMediaSourceMethods.getText(beat.image.code, context);
@@ -3,7 +3,7 @@ import { MulmoMediaSourceMethods } from "../../methods/mulmo_media_source.js";
3
3
  export const processSource = (imageType) => {
4
4
  return (params) => {
5
5
  const { beat, context } = params;
6
- if (!beat || !beat.image || beat.image.type !== imageType)
6
+ if (!beat?.image || beat.image.type !== imageType)
7
7
  return;
8
8
  const path = MulmoMediaSourceMethods.resolve(beat.image.source, context);
9
9
  if (path) {
@@ -1,5 +1,5 @@
1
1
  import { MulmoStudio, MulmoScript, MulmoPresentationStyle } from "../types/index.js";
2
- export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, currentStudio: MulmoStudio | undefined, fileName: string, videoCaption?: string, presentationStyle?: MulmoPresentationStyle | null) => {
2
+ export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, currentStudio: MulmoStudio | undefined, fileName: string, videoCaptionLang?: string, presentationStyle?: MulmoPresentationStyle | null) => {
3
3
  beats: {
4
4
  duration?: number | undefined;
5
5
  startAt?: number | undefined;
@@ -16,6 +16,7 @@ export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, curre
16
16
  captionFile?: string | undefined;
17
17
  }[];
18
18
  script: {
19
+ lang: string;
19
20
  imageParams: {
20
21
  provider: string;
21
22
  model?: string | undefined;
@@ -312,7 +313,6 @@ export declare const createOrUpdateStudioData: (_mulmoScript: MulmoScript, curre
312
313
  } | undefined;
313
314
  enableLipSync?: boolean | undefined;
314
315
  }[];
315
- lang?: string | undefined;
316
316
  title?: string | undefined;
317
317
  description?: string | undefined;
318
318
  lipSyncParams?: {
@@ -37,7 +37,7 @@ const mulmoCredit = (speaker) => {
37
37
  },
38
38
  };
39
39
  };
40
- export const createOrUpdateStudioData = (_mulmoScript, currentStudio, fileName, videoCaption, presentationStyle) => {
40
+ export const createOrUpdateStudioData = (_mulmoScript, currentStudio, fileName, videoCaptionLang, presentationStyle) => {
41
41
  const mulmoScript = _mulmoScript.__test_invalid__ ? _mulmoScript : MulmoScriptMethods.validate(_mulmoScript); // validate and insert default value
42
42
  const studio = rebuildStudio(currentStudio, mulmoScript, fileName);
43
43
  // TODO: Move this code out of this function later
@@ -48,10 +48,10 @@ export const createOrUpdateStudioData = (_mulmoScript, currentStudio, fileName,
48
48
  }
49
49
  studio.script = MulmoScriptMethods.validate(mulmoScript); // update the script
50
50
  studio.beats = studio.script.beats.map((_, index) => studio.beats[index] ?? {});
51
- if (videoCaption) {
51
+ if (videoCaptionLang) {
52
52
  studio.script.captionParams = mulmoCaptionParamsSchema.parse({
53
53
  ...(studio.script.captionParams ?? {}),
54
- lang: videoCaption,
54
+ lang: videoCaptionLang,
55
55
  });
56
56
  }
57
57
  return studio;
@@ -67,9 +67,10 @@ export declare const provider2LipSyncAgent: {
67
67
  defaultModel: ReplicateModel;
68
68
  models: ReplicateModel[];
69
69
  modelParams: Record<ReplicateModel, {
70
- identifier?: `${string}/${string}:${string}`;
71
- video: string;
70
+ identifier?: `${string}/${string}:${string}` | `${string}/${string}`;
71
+ video?: string;
72
72
  audio: string;
73
+ image?: string;
73
74
  }>;
74
75
  };
75
76
  };
@@ -50,6 +50,8 @@ export const provider2MovieAgent = {
50
50
  "minimax/video-01",
51
51
  "minimax/hailuo-02",
52
52
  "pixverse/pixverse-v4.5",
53
+ "wan-video/wan-2.2-i2v-480p-fast",
54
+ "wan-video/wan-2.2-t2v-480p-fast",
53
55
  ],
54
56
  modelParams: {
55
57
  "bytedance/seedance-1-lite": {
@@ -110,6 +112,16 @@ export const provider2MovieAgent = {
110
112
  last_image: "last_frame_image",
111
113
  price_per_sec: 0.12,
112
114
  },
115
+ "wan-video/wan-2.2-i2v-480p-fast": {
116
+ durations: [5],
117
+ start_image: "image",
118
+ price_per_sec: 0.012,
119
+ },
120
+ "wan-video/wan-2.2-t2v-480p-fast": {
121
+ durations: [5],
122
+ start_image: undefined,
123
+ price_per_sec: 0.012,
124
+ },
113
125
  },
114
126
  },
115
127
  google: {
@@ -133,8 +145,8 @@ export const provider2SoundEffectAgent = {
133
145
  export const provider2LipSyncAgent = {
134
146
  replicate: {
135
147
  agentName: "lipSyncReplicateAgent",
136
- defaultModel: "bytedance/latentsync",
137
- models: ["bytedance/latentsync", "tmappdev/lipsync"],
148
+ defaultModel: "bytedance/omni-human",
149
+ models: ["bytedance/latentsync", "tmappdev/lipsync", "bytedance/omni-human"],
138
150
  modelParams: {
139
151
  "bytedance/latentsync": {
140
152
  identifier: "bytedance/latentsync:637ce1919f807ca20da3a448ddc2743535d2853649574cd52a933120e9b9e293",
@@ -146,6 +158,12 @@ export const provider2LipSyncAgent = {
146
158
  video: "video_input",
147
159
  audio: "audio_input",
148
160
  },
161
+ "bytedance/omni-human": {
162
+ identifier: "bytedance/omni-human",
163
+ image: "image",
164
+ audio: "audio",
165
+ price_per_sec: 0.14,
166
+ },
149
167
  /* NOTE: This model does not work with large base64 urls.
150
168
  "sync/lipsync-2": {
151
169
  video: "video",
@@ -4,6 +4,6 @@ interface Replacement {
4
4
  from: string;
5
5
  to: string;
6
6
  }
7
- export declare function replacePairsJa(str: string, replacements: Replacement[]): string;
7
+ export declare function replacePairsJa(replacements: Replacement[]): (str: string) => string;
8
8
  export declare const replacementsJa: Replacement[];
9
9
  export {};
@@ -24,14 +24,16 @@ export const recursiveSplitJa = (text) => {
24
24
  }, [text])
25
25
  .flat(1);
26
26
  };
27
- export function replacePairsJa(str, replacements) {
28
- return replacements.reduce((tmp, current) => {
29
- const { from, to } = current;
30
- // Escape any special regex characters in the 'from' string.
31
- const escapedFrom = from.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
32
- const regex = new RegExp(escapedFrom, "g");
33
- return tmp.replace(regex, to);
34
- }, str);
27
+ export function replacePairsJa(replacements) {
28
+ return (str) => {
29
+ return replacements.reduce((tmp, current) => {
30
+ const { from, to } = current;
31
+ // Escape any special regex characters in the 'from' string.
32
+ const escapedFrom = from.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
33
+ const regex = new RegExp(escapedFrom, "g");
34
+ return tmp.replace(regex, to);
35
+ }, str);
36
+ };
35
37
  }
36
38
  export const replacementsJa = [
37
39
  { from: "Anthropic", to: "アンスロピック" },
@@ -51,4 +53,6 @@ export const replacementsJa = [
51
53
  { from: "5つ", to: "いつつ" },
52
54
  { from: "危険な面", to: "危険なめん" },
53
55
  { from: "その通り!", to: "その通り。" },
56
+ { from: "%", to: "パーセント" },
57
+ { from: "IPO", to: "アイピーオー" },
54
58
  ];
@@ -21,12 +21,8 @@ export const text2hash = (input) => {
21
21
  return crypto.createHash("sha256").update(input).digest("hex");
22
22
  };
23
23
  export const localizedText = (beat, multiLingualData, lang) => {
24
- if (lang &&
25
- multiLingualData &&
26
- multiLingualData?.multiLingualTexts &&
27
- multiLingualData?.multiLingualTexts[lang] &&
28
- multiLingualData?.multiLingualTexts[lang].text) {
29
- return multiLingualData?.multiLingualTexts[lang].text;
24
+ if (lang && multiLingualData?.multiLingualTexts?.[lang]?.text) {
25
+ return multiLingualData.multiLingualTexts[lang].text;
30
26
  }
31
27
  return beat.text;
32
28
  };
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "mulmocast",
3
- "version": "1.1.5",
3
+ "version": "1.1.7",
4
4
  "description": "",
5
5
  "type": "module",
6
6
  "main": "lib/index.node.js",
@@ -105,7 +105,7 @@
105
105
  "prettier": "^3.6.2",
106
106
  "ts-node": "^10.9.2",
107
107
  "tsx": "^4.20.3",
108
- "typescript": "^5.7.3",
108
+ "typescript": "^5.9.2",
109
109
  "typescript-eslint": "^8.37.0"
110
110
  },
111
111
  "engines": {
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "[TITLE OF THE PRESENTAITON OR STORY]",
6
7
  "imageParams": {
7
8
  "images": {
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "Voice Over Test",
6
7
  "captionParams": {
7
8
  "lang": "en"
@@ -3,6 +3,7 @@
3
3
  "version": "1.1",
4
4
  "credit": "closing"
5
5
  },
6
+ "lang": "en",
6
7
  "title": "Testing OpenAI's new Image Generation",
7
8
  "description": "Hello",
8
9
  "imageParams": {
@@ -3,6 +3,7 @@
3
3
  "version": "1.1",
4
4
  "credit": "closing"
5
5
  },
6
+ "lang": "en",
6
7
  "title": "The Honey Trap vs. The Frontier: Engineering at Tesla and SpaceX",
7
8
  "description": "We dive into Elon Musk's philosophy on building environments where engineers can truly flourish, contrasting the comfort-focused 'honey trap' with the high-expectation culture at Tesla and SpaceX.",
8
9
  "speechParams": {
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "Media Test",
6
7
  "audioParams": {
7
8
  "introPadding": 0,
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "Audio Instructions Test",
6
7
  "speechParams": {
7
8
  "speakers": {
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "Beat Reference Test",
6
7
  "beats": [
7
8
  {
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "captionParams": {
6
7
  "lang": "ja",
7
8
  "styles": ["color: yellow"]
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "ElevenLabs Model Test",
6
7
  "references": [
7
8
  {
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "beats": [
6
7
  {
7
8
  "text": "Hello World",
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "speechParams": {
6
7
  "speakers": {
7
8
  "Presenter": {
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "ja",
5
6
  "title": "Media Test",
6
7
  "beats": [
7
8
  {
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "Test Image References",
6
7
  "imageParams": {
7
8
  "style": "Ghibli-style",
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "Test Images",
6
7
  "imageParams": {
7
8
  "style": "Photorealistic-style",
@@ -2,7 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
- "lang": "ja",
5
+ "lang": "en",
6
6
  "speechParams": {
7
7
  "speakers": {
8
8
  "Presenter": {
@@ -19,13 +19,69 @@
19
19
  },
20
20
  "beats": [
21
21
  {
22
- "text": "こんにちは",
22
+ "text": "Hello World",
23
23
  "image": {
24
24
  "type": "textSlide",
25
25
  "slide": {
26
26
  "title": "Hello World"
27
27
  }
28
28
  }
29
+ },
30
+ {
31
+ "text": "Switching Language and Switching Screen should match.",
32
+ "image": {
33
+ "type": "textSlide",
34
+ "slide": {
35
+ "title": "Switching Language"
36
+ }
37
+ }
38
+ },
39
+ {
40
+ "text": "May the force be with you.",
41
+ "image": {
42
+ "type": "textSlide",
43
+ "slide": {
44
+ "title": "May the force be with you"
45
+ }
46
+ }
47
+ },
48
+ {
49
+ "text": "Switching Language and Switching Screen should match even with a short duration.",
50
+ "duration": 1,
51
+ "image": {
52
+ "type": "textSlide",
53
+ "slide": {
54
+ "title": "Switching Language (short duration)"
55
+ }
56
+ }
57
+ },
58
+ {
59
+ "text": "Switching Language and Switching Screen should match even with a long duration.",
60
+ "duration": 8,
61
+ "image": {
62
+ "type": "textSlide",
63
+ "slide": {
64
+ "title": "Switching Language (long duration)"
65
+ }
66
+ }
67
+ },
68
+ {
69
+ "text": "The yield of TSMC's 3nm process exceeds 70%.",
70
+ "image": {
71
+ "type": "textSlide",
72
+ "slide": {
73
+ "title": "Text replacement test for nijivoice"
74
+ }
75
+ }
76
+ },
77
+ {
78
+ "text": "Goodbye",
79
+ "image": {
80
+ "type": "textSlide",
81
+ "slide": {
82
+ "title": "Goodbye"
83
+ }
84
+ }
29
85
  }
30
86
  ]
31
87
  }
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "Layout Test",
6
7
  "audioParams": {
7
8
  "introPadding": 0,
@@ -33,6 +33,15 @@
33
33
  "model": "tmappdev/lipsync"
34
34
  }
35
35
  },
36
+ {
37
+ "id": "future_omni_human_photo_realistic",
38
+ "text": "In the future, advancements in AI could revolutionize industries like healthcare, education, and transportation.",
39
+ "imagePrompt": "A female presenter is standing in front of a futuristic cityscape with AI-powered hospital and an autonomous vehicle. Photo realistic.",
40
+ "enableLipSync": true,
41
+ "lipSyncParams": {
42
+ "model": "bytedance/omni-human"
43
+ }
44
+ },
36
45
  {
37
46
  "id": "future_possibilities",
38
47
  "text": "In the future, advancements in AI could revolutionize industries like healthcare, education, and transportation.",
@@ -3,6 +3,7 @@
3
3
  "version": "1.1",
4
4
  "credit": "closing"
5
5
  },
6
+ "lang": "en",
6
7
  "audioParams": {
7
8
  "bgm": {
8
9
  "kind": "url",
@@ -2,6 +2,7 @@
2
2
  "$mulmocast": {
3
3
  "version": "1.1"
4
4
  },
5
+ "lang": "en",
5
6
  "title": "Media Test",
6
7
  "references": [
7
8
  {