mulmocast 2.1.22 → 2.1.24
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.
- package/lib/agents/image_genai_agent.js +14 -6
- package/lib/agents/image_openai_agent.js +4 -3
- package/lib/agents/movie_genai_agent.js +35 -19
- package/lib/agents/tts_openai_agent.js +4 -3
- package/lib/methods/mulmo_presentation_style.d.ts +2 -0
- package/lib/types/agent.d.ts +5 -0
- package/lib/types/provider2agent.d.ts +2 -1
- package/lib/types/provider2agent.js +9 -2
- package/lib/types/schema.d.ts +92 -0
- package/lib/types/schema.js +8 -0
- package/lib/utils/context.d.ts +42 -0
- package/lib/utils/openai_client.d.ts +17 -0
- package/lib/utils/openai_client.js +35 -0
- package/lib/utils/utils.js +3 -0
- package/package.json +5 -5
- package/scripts/test/test_audio_azure.json +24 -0
- package/scripts/test/test_images_azure.json +26 -0
- package/scripts/test/test_vertexai.json +21 -0
- package/lib/agents/tts_nijivoice_agent.d.ts +0 -5
- package/lib/agents/tts_nijivoice_agent.js +0 -76
- package/lib/agents/utils.d.ts +0 -1
- package/lib/agents/utils.js +0 -1
- package/lib/utils/const.d.ts +0 -15
- package/lib/utils/const.js +0 -15
- package/lib/utils/provider2agent.d.ts +0 -191
- package/lib/utils/provider2agent.js +0 -326
|
@@ -1,326 +0,0 @@
|
|
|
1
|
-
// node & browser
|
|
2
|
-
export const provider2TTSAgent = {
|
|
3
|
-
nijivoice: {
|
|
4
|
-
agentName: "ttsNijivoiceAgent",
|
|
5
|
-
hasLimitedConcurrency: true,
|
|
6
|
-
keyName: "NIJIVOICE_API_KEY",
|
|
7
|
-
},
|
|
8
|
-
openai: {
|
|
9
|
-
agentName: "ttsOpenaiAgent",
|
|
10
|
-
hasLimitedConcurrency: false,
|
|
11
|
-
defaultModel: "gpt-4o-mini-tts",
|
|
12
|
-
defaultVoice: "shimmer",
|
|
13
|
-
keyName: "OPENAI_API_KEY",
|
|
14
|
-
baseURLKeyName: "OPENAI_BASE_URL",
|
|
15
|
-
},
|
|
16
|
-
google: {
|
|
17
|
-
agentName: "ttsGoogleAgent",
|
|
18
|
-
hasLimitedConcurrency: false,
|
|
19
|
-
keyName: "GEMINI_API_KEY",
|
|
20
|
-
},
|
|
21
|
-
gemini: {
|
|
22
|
-
agentName: "ttsGeminiAgent",
|
|
23
|
-
hasLimitedConcurrency: false,
|
|
24
|
-
defaultModel: "gemini-2.5-flash-preview-tts",
|
|
25
|
-
defaultVoice: "Kore",
|
|
26
|
-
models: ["gemini-2.5-flash-preview-tts", "gemini-2.5-pro-preview-tts"],
|
|
27
|
-
keyName: "GEMINI_API_KEY",
|
|
28
|
-
},
|
|
29
|
-
elevenlabs: {
|
|
30
|
-
agentName: "ttsElevenlabsAgent",
|
|
31
|
-
hasLimitedConcurrency: true,
|
|
32
|
-
defaultModel: "eleven_multilingual_v2",
|
|
33
|
-
// Models | ElevenLabs Documentation
|
|
34
|
-
// https://elevenlabs.io/docs/models
|
|
35
|
-
models: ["eleven_multilingual_v2", "eleven_turbo_v2_5", "eleven_turbo_v2", "eleven_flash_v2_5", "eleven_flash_v2"],
|
|
36
|
-
keyName: "ELEVENLABS_API_KEY",
|
|
37
|
-
},
|
|
38
|
-
kotodama: {
|
|
39
|
-
agentName: "ttsKotodamaAgent",
|
|
40
|
-
hasLimitedConcurrency: true,
|
|
41
|
-
defaultVoice: "Atla",
|
|
42
|
-
defaultDecoration: "neutral",
|
|
43
|
-
keyName: "KOTODAMA_API_KEY",
|
|
44
|
-
},
|
|
45
|
-
mock: {
|
|
46
|
-
agentName: "mediaMockAgent",
|
|
47
|
-
hasLimitedConcurrency: true,
|
|
48
|
-
defaultModel: "mock-model",
|
|
49
|
-
models: ["mock-model"],
|
|
50
|
-
},
|
|
51
|
-
};
|
|
52
|
-
export const gptImages = ["gpt-image-1.5", "gpt-image-1", "gpt-image-1-mini"];
|
|
53
|
-
export const provider2ImageAgent = {
|
|
54
|
-
openai: {
|
|
55
|
-
agentName: "imageOpenaiAgent",
|
|
56
|
-
defaultModel: "gpt-image-1",
|
|
57
|
-
models: ["dall-e-3", ...gptImages],
|
|
58
|
-
keyName: "OPENAI_API_KEY",
|
|
59
|
-
baseURLKeyName: "OPENAI_BASE_URL",
|
|
60
|
-
},
|
|
61
|
-
google: {
|
|
62
|
-
agentName: "imageGenAIAgent",
|
|
63
|
-
defaultModel: "gemini-2.5-flash-image",
|
|
64
|
-
models: ["imagen-4.0-generate-preview-06-06", "imagen-4.0-ultra-generate-preview-06-06", "gemini-2.5-flash-image", "gemini-3-pro-image-preview"],
|
|
65
|
-
keyName: "GEMINI_API_KEY",
|
|
66
|
-
},
|
|
67
|
-
replicate: {
|
|
68
|
-
agentName: "imageReplicateAgent",
|
|
69
|
-
defaultModel: "bytedance/seedream-4",
|
|
70
|
-
models: ["bytedance/seedream-4", "qwen/qwen-image"],
|
|
71
|
-
keyName: "REPLICATE_API_TOKEN",
|
|
72
|
-
},
|
|
73
|
-
mock: {
|
|
74
|
-
agentName: "mediaMockAgent",
|
|
75
|
-
defaultModel: "mock-model",
|
|
76
|
-
models: ["mock-model"],
|
|
77
|
-
keyName: "",
|
|
78
|
-
},
|
|
79
|
-
};
|
|
80
|
-
export const provider2MovieAgent = {
|
|
81
|
-
replicate: {
|
|
82
|
-
agentName: "movieReplicateAgent",
|
|
83
|
-
defaultModel: "bytedance/seedance-1-lite",
|
|
84
|
-
keyName: "REPLICATE_API_TOKEN",
|
|
85
|
-
models: [
|
|
86
|
-
"bytedance/seedance-1-lite",
|
|
87
|
-
"bytedance/seedance-1-pro",
|
|
88
|
-
"kwaivgi/kling-v1.6-pro",
|
|
89
|
-
"kwaivgi/kling-v2.1",
|
|
90
|
-
"kwaivgi/kling-v2.1-master",
|
|
91
|
-
"google/veo-2",
|
|
92
|
-
"google/veo-3",
|
|
93
|
-
"google/veo-3-fast",
|
|
94
|
-
"minimax/video-01",
|
|
95
|
-
"minimax/hailuo-02",
|
|
96
|
-
"minimax/hailuo-02-fast",
|
|
97
|
-
"pixverse/pixverse-v4.5",
|
|
98
|
-
"wan-video/wan-2.2-i2v-fast",
|
|
99
|
-
"wan-video/wan-2.2-t2v-fast",
|
|
100
|
-
],
|
|
101
|
-
modelParams: {
|
|
102
|
-
"bytedance/seedance-1-lite": {
|
|
103
|
-
durations: [5, 10],
|
|
104
|
-
start_image: "image",
|
|
105
|
-
last_image: "last_frame_image",
|
|
106
|
-
price_per_sec: 0.036, // in USD
|
|
107
|
-
},
|
|
108
|
-
"bytedance/seedance-1-pro": {
|
|
109
|
-
durations: [5, 10],
|
|
110
|
-
start_image: "image",
|
|
111
|
-
last_image: "last_frame_image",
|
|
112
|
-
price_per_sec: 0.15,
|
|
113
|
-
},
|
|
114
|
-
"kwaivgi/kling-v1.6-pro": {
|
|
115
|
-
durations: [5, 10],
|
|
116
|
-
start_image: "start_image",
|
|
117
|
-
price_per_sec: 0.095,
|
|
118
|
-
},
|
|
119
|
-
"kwaivgi/kling-v2.1": {
|
|
120
|
-
durations: [5, 10],
|
|
121
|
-
start_image: "start_image",
|
|
122
|
-
price_per_sec: 0.05,
|
|
123
|
-
},
|
|
124
|
-
"kwaivgi/kling-v2.1-master": {
|
|
125
|
-
durations: [5, 10],
|
|
126
|
-
start_image: "start_image",
|
|
127
|
-
price_per_sec: 0.28,
|
|
128
|
-
},
|
|
129
|
-
"google/veo-2": {
|
|
130
|
-
durations: [5, 6, 7, 8],
|
|
131
|
-
start_image: "image",
|
|
132
|
-
price_per_sec: 0.5,
|
|
133
|
-
},
|
|
134
|
-
"google/veo-3": {
|
|
135
|
-
durations: [8],
|
|
136
|
-
start_image: "image",
|
|
137
|
-
price_per_sec: 0.75,
|
|
138
|
-
},
|
|
139
|
-
"google/veo-3-fast": {
|
|
140
|
-
durations: [8],
|
|
141
|
-
start_image: "image",
|
|
142
|
-
price_per_sec: 0.4,
|
|
143
|
-
},
|
|
144
|
-
"minimax/video-01": {
|
|
145
|
-
durations: [6],
|
|
146
|
-
start_image: "first_frame_image",
|
|
147
|
-
price_per_sec: 0.5,
|
|
148
|
-
},
|
|
149
|
-
"minimax/hailuo-02": {
|
|
150
|
-
durations: [6], // NOTE: 10 for only 720p
|
|
151
|
-
start_image: "first_frame_image",
|
|
152
|
-
price_per_sec: 0.08,
|
|
153
|
-
},
|
|
154
|
-
"minimax/hailuo-02-fast": {
|
|
155
|
-
durations: [6, 10], // NOTE: 512P
|
|
156
|
-
start_image: "first_frame_image",
|
|
157
|
-
price_per_sec: 0.0166,
|
|
158
|
-
},
|
|
159
|
-
"pixverse/pixverse-v4.5": {
|
|
160
|
-
durations: [5, 8],
|
|
161
|
-
start_image: "image",
|
|
162
|
-
last_image: "last_frame_image",
|
|
163
|
-
price_per_sec: 0.12,
|
|
164
|
-
},
|
|
165
|
-
"wan-video/wan-2.2-i2v-fast": {
|
|
166
|
-
durations: [5],
|
|
167
|
-
start_image: "image",
|
|
168
|
-
price_per_sec: 0.012,
|
|
169
|
-
},
|
|
170
|
-
"wan-video/wan-2.2-t2v-fast": {
|
|
171
|
-
durations: [5],
|
|
172
|
-
start_image: undefined,
|
|
173
|
-
price_per_sec: 0.012,
|
|
174
|
-
},
|
|
175
|
-
},
|
|
176
|
-
},
|
|
177
|
-
google: {
|
|
178
|
-
agentName: "movieGenAIAgent",
|
|
179
|
-
defaultModel: "veo-2.0-generate-001",
|
|
180
|
-
models: ["veo-2.0-generate-001", "veo-3.0-generate-001", "veo-3.1-generate-preview"],
|
|
181
|
-
keyName: "GEMINI_API_KEY",
|
|
182
|
-
modelParams: {
|
|
183
|
-
"veo-3.1-generate-preview": {
|
|
184
|
-
durations: [4, 6, 8],
|
|
185
|
-
},
|
|
186
|
-
"veo-3.0-generate-001": {
|
|
187
|
-
durations: [4, 6, 8],
|
|
188
|
-
},
|
|
189
|
-
"veo-2.0-generate-001": {
|
|
190
|
-
durations: [5, 6, 7, 8],
|
|
191
|
-
},
|
|
192
|
-
},
|
|
193
|
-
},
|
|
194
|
-
mock: {
|
|
195
|
-
agentName: "mediaMockAgent",
|
|
196
|
-
defaultModel: "mock-model",
|
|
197
|
-
models: ["mock-model"],
|
|
198
|
-
keyName: "",
|
|
199
|
-
modelParams: {},
|
|
200
|
-
},
|
|
201
|
-
};
|
|
202
|
-
export const provider2SoundEffectAgent = {
|
|
203
|
-
replicate: {
|
|
204
|
-
agentName: "soundEffectReplicateAgent",
|
|
205
|
-
defaultModel: "zsxkib/mmaudio",
|
|
206
|
-
keyName: "REPLICATE_API_TOKEN",
|
|
207
|
-
models: ["zsxkib/mmaudio"],
|
|
208
|
-
modelParams: {
|
|
209
|
-
"zsxkib/mmaudio": {
|
|
210
|
-
identifier: "zsxkib/mmaudio:62871fb59889b2d7c13777f08deb3b36bdff88f7e1d53a50ad7694548a41b484",
|
|
211
|
-
},
|
|
212
|
-
},
|
|
213
|
-
},
|
|
214
|
-
};
|
|
215
|
-
export const provider2LipSyncAgent = {
|
|
216
|
-
replicate: {
|
|
217
|
-
agentName: "lipSyncReplicateAgent",
|
|
218
|
-
defaultModel: "bytedance/omni-human",
|
|
219
|
-
keyName: "REPLICATE_API_TOKEN",
|
|
220
|
-
models: ["bytedance/latentsync", "tmappdev/lipsync", "bytedance/omni-human"],
|
|
221
|
-
modelParams: {
|
|
222
|
-
"bytedance/latentsync": {
|
|
223
|
-
identifier: "bytedance/latentsync:637ce1919f807ca20da3a448ddc2743535d2853649574cd52a933120e9b9e293",
|
|
224
|
-
video: "video",
|
|
225
|
-
audio: "audio",
|
|
226
|
-
},
|
|
227
|
-
"tmappdev/lipsync": {
|
|
228
|
-
identifier: "tmappdev/lipsync:c54ce2fe673ea59b857b91250b3d71a2cd304a78f2370687632805c8405fbf4c",
|
|
229
|
-
video: "video_input",
|
|
230
|
-
audio: "audio_input",
|
|
231
|
-
},
|
|
232
|
-
"bytedance/omni-human": {
|
|
233
|
-
identifier: "bytedance/omni-human",
|
|
234
|
-
image: "image",
|
|
235
|
-
audio: "audio",
|
|
236
|
-
price_per_sec: 0.14,
|
|
237
|
-
},
|
|
238
|
-
/* NOTE: This model does not work with large base64 urls.
|
|
239
|
-
"sync/lipsync-2": {
|
|
240
|
-
video: "video",
|
|
241
|
-
audio: "audio",
|
|
242
|
-
},
|
|
243
|
-
*/
|
|
244
|
-
/* NOTE: This model does not work well for some unknown reason.
|
|
245
|
-
"kwaivgi/kling-lip-sync": {
|
|
246
|
-
video: "video_url",
|
|
247
|
-
audio: "audio_file",
|
|
248
|
-
},
|
|
249
|
-
*/
|
|
250
|
-
},
|
|
251
|
-
},
|
|
252
|
-
};
|
|
253
|
-
// : Record<LLM, { agent: string; defaultModel: string; max_tokens: number }>
|
|
254
|
-
export const provider2LLMAgent = {
|
|
255
|
-
openai: {
|
|
256
|
-
agentName: "openAIAgent",
|
|
257
|
-
defaultModel: "gpt-5",
|
|
258
|
-
keyName: "OPENAI_API_KEY",
|
|
259
|
-
baseURLKeyName: "OPENAI_BASE_URL",
|
|
260
|
-
max_tokens: 8192,
|
|
261
|
-
models: [
|
|
262
|
-
"gpt-5",
|
|
263
|
-
"gpt-5-nano",
|
|
264
|
-
"gpt-5-mini",
|
|
265
|
-
"gpt-4.1",
|
|
266
|
-
"gpt-4.1-mini",
|
|
267
|
-
"gpt-4.1-nano",
|
|
268
|
-
"o3",
|
|
269
|
-
"o3-mini",
|
|
270
|
-
"o3-pro",
|
|
271
|
-
"o1",
|
|
272
|
-
"o1-pro",
|
|
273
|
-
"gpt-4o",
|
|
274
|
-
"gpt-4o-mini",
|
|
275
|
-
],
|
|
276
|
-
},
|
|
277
|
-
anthropic: {
|
|
278
|
-
agentName: "anthropicAgent",
|
|
279
|
-
defaultModel: "claude-3-7-sonnet-20250219",
|
|
280
|
-
max_tokens: 8192,
|
|
281
|
-
models: ["claude-opus-4-1-20250805", "claude-opus-4-20250514", "claude-sonnet-4-20250514", "claude-3-7-sonnet-20250219", "claude-3-haiku-20240307"],
|
|
282
|
-
keyName: "ANTHROPIC_API_KEY",
|
|
283
|
-
apiKeyNameOverride: "ANTHROPIC_API_TOKEN",
|
|
284
|
-
// GraphAI is currently using ANTHROPIC_API_KEY, but the official name is ANTHROPIC_API_TOKEN.
|
|
285
|
-
},
|
|
286
|
-
gemini: {
|
|
287
|
-
agentName: "geminiAgent",
|
|
288
|
-
defaultModel: "gemini-2.5-flash",
|
|
289
|
-
max_tokens: 8192,
|
|
290
|
-
models: ["gemini-2.5-pro", "gemini-2.5-flash", "gemini-2.5-flash-lite", "gemini-2.0-flash"],
|
|
291
|
-
keyName: "GEMINI_API_KEY",
|
|
292
|
-
},
|
|
293
|
-
groq: {
|
|
294
|
-
agentName: "groqAgent",
|
|
295
|
-
defaultModel: "llama-3.1-8b-instant",
|
|
296
|
-
keyName: "GROQ_API_KEY",
|
|
297
|
-
max_tokens: 4096,
|
|
298
|
-
models: ["llama-3.1-8b-instant", "llama-3.3-70b-versatile", "deepseek-r1-distill-llama-70b", "openai/gpt-oss-120b", "openai/gpt-oss-20b"],
|
|
299
|
-
},
|
|
300
|
-
mock: {
|
|
301
|
-
agentName: "mediaMockAgent",
|
|
302
|
-
defaultModel: "mock",
|
|
303
|
-
max_tokens: 4096,
|
|
304
|
-
models: ["mock"],
|
|
305
|
-
},
|
|
306
|
-
};
|
|
307
|
-
export const defaultProviders = {
|
|
308
|
-
tts: "openai",
|
|
309
|
-
text2image: "openai",
|
|
310
|
-
text2movie: "replicate",
|
|
311
|
-
text2Html: "openai",
|
|
312
|
-
llm: "openai",
|
|
313
|
-
soundEffect: "replicate",
|
|
314
|
-
lipSync: "replicate",
|
|
315
|
-
};
|
|
316
|
-
export const llm = Object.keys(provider2LLMAgent);
|
|
317
|
-
export const htmlLLMProvider = ["openai", "anthropic", "mock"];
|
|
318
|
-
export const getModelDuration = (provider, model, movieDuration) => {
|
|
319
|
-
const modelParams = provider2MovieAgent[provider]?.modelParams;
|
|
320
|
-
const { durations } = modelParams[model];
|
|
321
|
-
if (durations && movieDuration) {
|
|
322
|
-
const largerDurations = durations.filter((d) => d >= movieDuration);
|
|
323
|
-
return largerDurations.length > 0 ? largerDurations[0] : durations[durations.length - 1];
|
|
324
|
-
}
|
|
325
|
-
return durations?.[0];
|
|
326
|
-
};
|