ai 7.0.7 → 7.0.8
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/CHANGELOG.md +10 -0
- package/dist/index.d.ts +19 -2
- package/dist/index.js +62 -44
- package/dist/index.js.map +1 -1
- package/dist/internal/index.js +1 -1
- package/docs/03-ai-sdk-core/38-video-generation.mdx +40 -0
- package/docs/07-reference/01-ai-sdk-core/13-generate-video.mdx +12 -0
- package/package.json +4 -4
- package/src/generate-video/generate-video.ts +122 -49
package/dist/internal/index.js
CHANGED
|
@@ -171,6 +171,46 @@ const { video } = await generateVideo({
|
|
|
171
171
|
});
|
|
172
172
|
```
|
|
173
173
|
|
|
174
|
+
### First and Last Frame
|
|
175
|
+
|
|
176
|
+
Some video models support first-last-frame generation, where you provide the
|
|
177
|
+
starting and/or ending frames of the video. Use the `frameImages` option to pass
|
|
178
|
+
role-tagged images in a provider-agnostic way:
|
|
179
|
+
|
|
180
|
+
```tsx highlight={"5-16"}
|
|
181
|
+
const { video } = await generateVideo({
|
|
182
|
+
model: __VIDEO_MODEL__,
|
|
183
|
+
prompt: 'The cat walks across the scene and transforms into a dog by the end',
|
|
184
|
+
frameImages: [
|
|
185
|
+
{
|
|
186
|
+
image: 'https://example.com/first-frame.png',
|
|
187
|
+
frameType: 'first_frame',
|
|
188
|
+
},
|
|
189
|
+
{
|
|
190
|
+
image: 'https://example.com/last-frame.png',
|
|
191
|
+
frameType: 'last_frame',
|
|
192
|
+
},
|
|
193
|
+
],
|
|
194
|
+
});
|
|
195
|
+
```
|
|
196
|
+
|
|
197
|
+
### Reference Images
|
|
198
|
+
|
|
199
|
+
Some video models support reference-to-video generation, where you provide one or
|
|
200
|
+
more reference images that the model incorporates into the generated video. Use the `inputReferences` option to pass
|
|
201
|
+
these images in a provider-agnostic way:
|
|
202
|
+
|
|
203
|
+
```tsx highlight={"5-8"}
|
|
204
|
+
const { video } = await generateVideo({
|
|
205
|
+
model: __VIDEO_MODEL__,
|
|
206
|
+
prompt: 'The two characters meet in a bustling market',
|
|
207
|
+
inputReferences: [
|
|
208
|
+
'https://example.com/character-1.png',
|
|
209
|
+
'https://example.com/character-2.png',
|
|
210
|
+
],
|
|
211
|
+
});
|
|
212
|
+
```
|
|
213
|
+
|
|
174
214
|
### Providing a Seed
|
|
175
215
|
|
|
176
216
|
You can provide a seed to the `experimental_generateVideo` function to control the output of the video generation process.
|
|
@@ -110,6 +110,18 @@ console.log(videos);
|
|
|
110
110
|
isOptional: true,
|
|
111
111
|
description: 'Seed for the video generation.',
|
|
112
112
|
},
|
|
113
|
+
{
|
|
114
|
+
name: 'frameImages',
|
|
115
|
+
type: 'Array<{ image: DataContent; frameType: "first_frame" | "last_frame" }>',
|
|
116
|
+
isOptional: true,
|
|
117
|
+
description: 'Role-tagged image inputs for first-last-frame generation.',
|
|
118
|
+
},
|
|
119
|
+
{
|
|
120
|
+
name: 'inputReferences',
|
|
121
|
+
type: 'Array<DataContent>',
|
|
122
|
+
isOptional: true,
|
|
123
|
+
description: 'Reference image inputs for reference-to-video generation.',
|
|
124
|
+
},
|
|
113
125
|
{
|
|
114
126
|
name: 'generateAudio',
|
|
115
127
|
type: 'boolean',
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "ai",
|
|
3
|
-
"version": "7.0.
|
|
3
|
+
"version": "7.0.8",
|
|
4
4
|
"type": "module",
|
|
5
5
|
"description": "AI SDK by Vercel - build apps like ChatGPT, Claude, Gemini, and more with a single interface for any model using the Vercel AI Gateway or go direct to OpenAI, Anthropic, Google, or any other model provider.",
|
|
6
6
|
"license": "Apache-2.0",
|
|
@@ -42,9 +42,9 @@
|
|
|
42
42
|
}
|
|
43
43
|
},
|
|
44
44
|
"dependencies": {
|
|
45
|
-
"@ai-sdk/gateway": "4.0.
|
|
46
|
-
"@ai-sdk/provider": "4.0.
|
|
47
|
-
"@ai-sdk/provider-utils": "5.0.
|
|
45
|
+
"@ai-sdk/gateway": "4.0.6",
|
|
46
|
+
"@ai-sdk/provider": "4.0.1",
|
|
47
|
+
"@ai-sdk/provider-utils": "5.0.2"
|
|
48
48
|
},
|
|
49
49
|
"devDependencies": {
|
|
50
50
|
"@edge-runtime/vm": "^5.0.0",
|
|
@@ -2,6 +2,8 @@ import type {
|
|
|
2
2
|
Experimental_VideoModelV4,
|
|
3
3
|
Experimental_VideoModelV4CallOptions,
|
|
4
4
|
Experimental_VideoModelV4File,
|
|
5
|
+
Experimental_VideoModelV4FrameImage,
|
|
6
|
+
Experimental_VideoModelV4FrameType,
|
|
5
7
|
SharedV4ProviderMetadata,
|
|
6
8
|
} from '@ai-sdk/provider';
|
|
7
9
|
import {
|
|
@@ -26,6 +28,7 @@ import { prepareRetries } from '../util/prepare-retries';
|
|
|
26
28
|
import { VERSION } from '../version';
|
|
27
29
|
import type { GenerateVideoResult } from './generate-video-result';
|
|
28
30
|
import { splitDataUrl } from '../prompt/split-data-url';
|
|
31
|
+
import { convertDataContentToUint8Array } from '../prompt/data-content';
|
|
29
32
|
|
|
30
33
|
export type GenerateVideoPrompt =
|
|
31
34
|
| string
|
|
@@ -45,6 +48,8 @@ export type GenerateVideoPrompt =
|
|
|
45
48
|
* @param duration - Duration of the video in seconds.
|
|
46
49
|
* @param fps - Frames per second for the video.
|
|
47
50
|
* @param seed - Seed for the video generation.
|
|
51
|
+
* @param frameImages - Role-tagged image inputs for image-to-video and first-last-frame generation.
|
|
52
|
+
* @param inputReferences - Reference image inputs for reference-to-video generation.
|
|
48
53
|
* @param generateAudio - Whether the model should generate audio alongside the video.
|
|
49
54
|
* @param providerOptions - Additional provider-specific options that are passed through to the provider
|
|
50
55
|
* as body parameters.
|
|
@@ -66,6 +71,8 @@ export async function experimental_generateVideo({
|
|
|
66
71
|
duration,
|
|
67
72
|
fps,
|
|
68
73
|
seed,
|
|
74
|
+
frameImages,
|
|
75
|
+
inputReferences,
|
|
69
76
|
generateAudio,
|
|
70
77
|
providerOptions,
|
|
71
78
|
maxRetries: maxRetriesArg,
|
|
@@ -118,6 +125,26 @@ export async function experimental_generateVideo({
|
|
|
118
125
|
*/
|
|
119
126
|
seed?: number;
|
|
120
127
|
|
|
128
|
+
/**
|
|
129
|
+
* Role-tagged image inputs for image-to-video and first-last-frame generation.
|
|
130
|
+
*/
|
|
131
|
+
frameImages?: Array<{
|
|
132
|
+
/**
|
|
133
|
+
* The image for this frame.
|
|
134
|
+
*/
|
|
135
|
+
image: DataContent;
|
|
136
|
+
|
|
137
|
+
/**
|
|
138
|
+
* Which frame this image represents.
|
|
139
|
+
*/
|
|
140
|
+
frameType: Experimental_VideoModelV4FrameType;
|
|
141
|
+
}>;
|
|
142
|
+
|
|
143
|
+
/**
|
|
144
|
+
* Reference image inputs for reference-to-video generation.
|
|
145
|
+
*/
|
|
146
|
+
inputReferences?: Array<DataContent>;
|
|
147
|
+
|
|
121
148
|
/**
|
|
122
149
|
* Whether the model should generate audio alongside the video.
|
|
123
150
|
*/
|
|
@@ -172,6 +199,55 @@ export async function experimental_generateVideo({
|
|
|
172
199
|
|
|
173
200
|
const { prompt, image } = normalizePrompt(promptArg);
|
|
174
201
|
|
|
202
|
+
const normalizedFrameImages:
|
|
203
|
+
| Array<Experimental_VideoModelV4FrameImage>
|
|
204
|
+
| undefined = frameImages?.map(frame => ({
|
|
205
|
+
image: normalizeImageData(frame.image),
|
|
206
|
+
frameType: frame.frameType,
|
|
207
|
+
}));
|
|
208
|
+
|
|
209
|
+
const normalizedInputReferences:
|
|
210
|
+
| Array<Experimental_VideoModelV4File>
|
|
211
|
+
| undefined = inputReferences?.map(reference =>
|
|
212
|
+
normalizeImageData(reference),
|
|
213
|
+
);
|
|
214
|
+
|
|
215
|
+
const effectiveInputReferences =
|
|
216
|
+
normalizedFrameImages != null && normalizedFrameImages.length > 0
|
|
217
|
+
? undefined
|
|
218
|
+
: normalizedInputReferences;
|
|
219
|
+
|
|
220
|
+
const warnings: Array<Warning> = [];
|
|
221
|
+
|
|
222
|
+
if (
|
|
223
|
+
normalizedFrameImages != null &&
|
|
224
|
+
normalizedFrameImages.length > 0 &&
|
|
225
|
+
normalizedInputReferences != null &&
|
|
226
|
+
normalizedInputReferences.length > 0
|
|
227
|
+
) {
|
|
228
|
+
warnings.push({
|
|
229
|
+
type: 'other',
|
|
230
|
+
message:
|
|
231
|
+
'inputReferences were ignored because frameImages were provided; ' +
|
|
232
|
+
'frameImages and inputReferences cannot be combined.',
|
|
233
|
+
});
|
|
234
|
+
}
|
|
235
|
+
|
|
236
|
+
const firstFrameImage = normalizedFrameImages?.find(
|
|
237
|
+
frame => frame.frameType === 'first_frame',
|
|
238
|
+
)?.image;
|
|
239
|
+
|
|
240
|
+
if (image != null && firstFrameImage != null) {
|
|
241
|
+
warnings.push({
|
|
242
|
+
type: 'other',
|
|
243
|
+
message:
|
|
244
|
+
'prompt.image was ignored because a first_frame frameImage was provided; ' +
|
|
245
|
+
'the first_frame frameImage takes precedence as the start image.',
|
|
246
|
+
});
|
|
247
|
+
}
|
|
248
|
+
|
|
249
|
+
const resolvedImage = firstFrameImage ?? image;
|
|
250
|
+
|
|
175
251
|
const maxVideosPerCallWithDefault =
|
|
176
252
|
maxVideosPerCall ?? (await invokeModelMaxVideosPerCall(model)) ?? 1;
|
|
177
253
|
|
|
@@ -194,8 +270,10 @@ export async function experimental_generateVideo({
|
|
|
194
270
|
duration,
|
|
195
271
|
fps,
|
|
196
272
|
seed,
|
|
273
|
+
image: resolvedImage,
|
|
274
|
+
frameImages: normalizedFrameImages,
|
|
275
|
+
inputReferences: effectiveInputReferences,
|
|
197
276
|
generateAudio,
|
|
198
|
-
image,
|
|
199
277
|
providerOptions: providerOptions ?? {},
|
|
200
278
|
headers: headersWithUserAgent,
|
|
201
279
|
abortSignal,
|
|
@@ -206,7 +284,6 @@ export async function experimental_generateVideo({
|
|
|
206
284
|
|
|
207
285
|
// collect result videos, warnings, and response metadata
|
|
208
286
|
const videos: Array<GeneratedFile> = [];
|
|
209
|
-
const warnings: Array<Warning> = [];
|
|
210
287
|
const responses: Array<VideoModelResponseMetadata> = [];
|
|
211
288
|
const providerMetadata: SharedV4ProviderMetadata = {};
|
|
212
289
|
|
|
@@ -342,59 +419,55 @@ function normalizePrompt(promptArg: GenerateVideoPrompt): {
|
|
|
342
419
|
};
|
|
343
420
|
}
|
|
344
421
|
|
|
345
|
-
|
|
346
|
-
|
|
347
|
-
|
|
348
|
-
|
|
349
|
-
|
|
350
|
-
|
|
351
|
-
|
|
352
|
-
|
|
353
|
-
|
|
354
|
-
|
|
355
|
-
|
|
356
|
-
|
|
357
|
-
|
|
358
|
-
|
|
359
|
-
|
|
360
|
-
|
|
361
|
-
|
|
362
|
-
|
|
363
|
-
|
|
364
|
-
|
|
365
|
-
|
|
366
|
-
|
|
367
|
-
|
|
368
|
-
|
|
369
|
-
|
|
370
|
-
|
|
371
|
-
|
|
372
|
-
|
|
373
|
-
|
|
374
|
-
image = {
|
|
375
|
-
type: 'file',
|
|
376
|
-
mediaType,
|
|
377
|
-
data: bytes,
|
|
378
|
-
};
|
|
379
|
-
}
|
|
380
|
-
} else if (dataContent instanceof Uint8Array) {
|
|
381
|
-
const mediaType =
|
|
382
|
-
detectMediaType({
|
|
383
|
-
data: dataContent,
|
|
384
|
-
topLevelType: 'image',
|
|
385
|
-
}) ?? 'image/png';
|
|
386
|
-
|
|
387
|
-
image = {
|
|
422
|
+
return {
|
|
423
|
+
prompt: promptArg.text,
|
|
424
|
+
image:
|
|
425
|
+
promptArg.image != null ? normalizeImageData(promptArg.image) : undefined,
|
|
426
|
+
};
|
|
427
|
+
}
|
|
428
|
+
|
|
429
|
+
/**
|
|
430
|
+
* Normalizes a {@link DataContent} image into a {@link Experimental_VideoModelV4File}.
|
|
431
|
+
* Accepts a URL string, a data URL, a base64 string, or binary image data.
|
|
432
|
+
*/
|
|
433
|
+
function normalizeImageData(
|
|
434
|
+
dataContent: DataContent,
|
|
435
|
+
): Experimental_VideoModelV4File {
|
|
436
|
+
if (typeof dataContent === 'string') {
|
|
437
|
+
if (
|
|
438
|
+
dataContent.startsWith('http://') ||
|
|
439
|
+
dataContent.startsWith('https://')
|
|
440
|
+
) {
|
|
441
|
+
return {
|
|
442
|
+
type: 'url',
|
|
443
|
+
url: dataContent,
|
|
444
|
+
};
|
|
445
|
+
}
|
|
446
|
+
|
|
447
|
+
if (dataContent.startsWith('data:')) {
|
|
448
|
+
const { mediaType, base64Content } = splitDataUrl(dataContent);
|
|
449
|
+
return {
|
|
388
450
|
type: 'file',
|
|
389
|
-
mediaType,
|
|
390
|
-
data:
|
|
451
|
+
mediaType: mediaType ?? 'image/png',
|
|
452
|
+
data: convertBase64ToUint8Array(base64Content ?? ''),
|
|
391
453
|
};
|
|
392
454
|
}
|
|
455
|
+
|
|
456
|
+
const bytes = convertBase64ToUint8Array(dataContent);
|
|
457
|
+
return {
|
|
458
|
+
type: 'file',
|
|
459
|
+
mediaType:
|
|
460
|
+
detectMediaType({ data: bytes, topLevelType: 'image' }) ?? 'image/png',
|
|
461
|
+
data: bytes,
|
|
462
|
+
};
|
|
393
463
|
}
|
|
394
464
|
|
|
465
|
+
const bytes = convertDataContentToUint8Array(dataContent);
|
|
395
466
|
return {
|
|
396
|
-
|
|
397
|
-
|
|
467
|
+
type: 'file',
|
|
468
|
+
mediaType:
|
|
469
|
+
detectMediaType({ data: bytes, topLevelType: 'image' }) ?? 'image/png',
|
|
470
|
+
data: bytes,
|
|
398
471
|
};
|
|
399
472
|
}
|
|
400
473
|
|