@sogni-ai/sogni-client 5.82.0 → 5.84.0
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/AGENTS.md +9 -2
- package/CHANGELOG.md +22 -0
- package/README.md +59 -6
- package/dist/Changelog/index.d.ts +28 -0
- package/dist/Changelog/index.js +81 -0
- package/dist/Changelog/index.js.map +1 -0
- package/dist/Changelog/types.d.ts +83 -0
- package/dist/Changelog/types.js +3 -0
- package/dist/Changelog/types.js.map +1 -0
- package/dist/Chat/_hostedToolsManifest.generated.js +15 -6
- package/dist/Chat/_hostedToolsManifest.generated.js.map +1 -1
- package/dist/Chat/hostedToolValidator.js +24 -2
- package/dist/Chat/hostedToolValidator.js.map +1 -1
- package/dist/Chat/tools.js +1 -1
- package/dist/Chat/tools.js.map +1 -1
- package/dist/Projects/createJobRequestMessage.js +49 -23
- package/dist/Projects/createJobRequestMessage.js.map +1 -1
- package/dist/Projects/types/index.d.ts +20 -13
- package/dist/Projects/types/index.js.map +1 -1
- package/dist/Projects/utils/index.d.ts +24 -6
- package/dist/Projects/utils/index.js +34 -9
- package/dist/Projects/utils/index.js.map +1 -1
- package/dist/index.d.ts +11 -2
- package/dist/index.js +16 -11
- package/dist/index.js.map +1 -1
- package/dist/lib/validation.js +1 -0
- package/dist/lib/validation.js.map +1 -1
- package/dist/version.d.ts +1 -1
- package/dist/version.js +1 -1
- package/dist-esm/Changelog/index.js +76 -0
- package/dist-esm/Changelog/index.js.map +1 -0
- package/dist-esm/Changelog/types.js +2 -0
- package/dist-esm/Changelog/types.js.map +1 -0
- package/dist-esm/Chat/_hostedToolsManifest.generated.js +15 -6
- package/dist-esm/Chat/_hostedToolsManifest.generated.js.map +1 -1
- package/dist-esm/Chat/hostedToolValidator.js +24 -2
- package/dist-esm/Chat/hostedToolValidator.js.map +1 -1
- package/dist-esm/Chat/tools.js +1 -1
- package/dist-esm/Chat/tools.js.map +1 -1
- package/dist-esm/Projects/createJobRequestMessage.js +50 -24
- package/dist-esm/Projects/createJobRequestMessage.js.map +1 -1
- package/dist-esm/Projects/types/index.js.map +1 -1
- package/dist-esm/Projects/utils/index.js +31 -8
- package/dist-esm/Projects/utils/index.js.map +1 -1
- package/dist-esm/index.js +4 -2
- package/dist-esm/index.js.map +1 -1
- package/dist-esm/lib/validation.js +1 -0
- package/dist-esm/lib/validation.js.map +1 -1
- package/dist-esm/version.js +1 -1
- package/llms-full.txt +74 -11
- package/llms.txt +46 -7
- package/package.json +6 -3
- package/src/Changelog/index.ts +75 -0
- package/src/Changelog/types.ts +96 -0
- package/src/Chat/_hostedToolsManifest.generated.ts +15 -6
- package/src/Chat/hostedToolValidator.ts +30 -2
- package/src/Chat/tools.ts +1 -1
- package/src/Projects/createJobRequestMessage.ts +54 -26
- package/src/Projects/types/index.ts +20 -13
- package/src/Projects/utils/index.ts +36 -8
- package/src/index.ts +24 -0
- package/src/lib/validation.ts +1 -0
- package/src/version.ts +1 -1
|
@@ -806,7 +806,7 @@ export const SOGNI_HOSTED_TOOLS_MANIFEST: unknown = {
|
|
|
806
806
|
"type": "function",
|
|
807
807
|
"function": {
|
|
808
808
|
"name": "upscale_video",
|
|
809
|
-
"description": "Upscale an existing video to 1080p or 1440p with FlashVSR
|
|
809
|
+
"description": "Upscale an existing video to 1080p, 1440p or 2160p (4K) with recommended RelayVSR, or to 1080p or 1440p with FlashVSR. This is promptless, deterministic enhancement, not a generative edit: it keeps every source frame, the exact frame rate, the full aspect ratio, and the original audio, and it never changes content, trims, crops, restyles, or interpolates frames. Use it when the user asks to upscale, enlarge, sharpen, or increase the resolution of an uploaded or previously generated video, or wants an HD, 1080p, 1440p, 2K, or 4K copy of it. Do not use it to create, restyle, extend, or edit a video; when the user explicitly asks a generative model such as Seedance to re-render the clip, use video_to_video. Explicitly set model to relayvsr for new upscales. Use flashvsr when the user requests FlashVSR or specifically wants its detailPreference, processingSpeed or seed controls. Omitted model retains FlashVSR for compatibility. With FlashVSR the output is at most twice the source size, so 1080p needs a source whose short edge is 540-768px and 1440p needs 720-768px, and the source can be at most about 1344x768 pixels (768x1344 in portrait). With RelayVSR the output is at most twice the source size for a source up to 1920x1080 pixels (1080x1920 in portrait), or four times for a source up to 1344x768 (768x1344 in portrait), so 2160p needs a source whose short edge is 540-768px or exactly 1080px, and the output's long edge can be at most 3840px. The source must also be 1-60 fps and at most 100 MB. The server sets the maximum clip length and rejects a source that is too long; never quote a length limit yourself, and if the tool reports that rejection, relay that error to the user. FlashVSR cannot produce 4K or any size above 1440p, so a 4K request always uses relayvsr. Nothing here produces a size above 2160p; when the user asks for one, say so and offer 2160p. Each upscale costs credits based on the source video's size and length. With FlashVSR, optional detailPreference (stable or sharper), processingSpeed (stable or faster), and seed tune the result; omit them unless the user asks, and none of them changes the price. RelayVSR refuses all three.",
|
|
810
810
|
"parameters": {
|
|
811
811
|
"type": "object",
|
|
812
812
|
"properties": {
|
|
@@ -814,13 +814,22 @@ export const SOGNI_HOSTED_TOOLS_MANIFEST: unknown = {
|
|
|
814
814
|
"type": "number",
|
|
815
815
|
"description": "Source video to upscale. Non-negative values select a prior generated video result by 0-based index. Negative values select uploaded videos: -1 is the first uploaded video, -2 the second, and so on. If omitted, use the latest generated video, falling back to the most recent uploaded video."
|
|
816
816
|
},
|
|
817
|
+
"model": {
|
|
818
|
+
"type": "string",
|
|
819
|
+
"enum": [
|
|
820
|
+
"flashvsr",
|
|
821
|
+
"relayvsr"
|
|
822
|
+
],
|
|
823
|
+
"description": "Upscaler. Explicitly select relayvsr, recommended for new upscales to 1080p, 1440p or 2160p. Use flashvsr when requested or for its detailPreference, processingSpeed or seed controls; it reaches 1080p or 1440p. RelayVSR takes none of those options. Omitted model retains FlashVSR for compatibility."
|
|
824
|
+
},
|
|
817
825
|
"targetResolution": {
|
|
818
826
|
"type": "number",
|
|
819
827
|
"enum": [
|
|
820
828
|
1080,
|
|
821
|
-
1440
|
|
829
|
+
1440,
|
|
830
|
+
2160
|
|
822
831
|
],
|
|
823
|
-
"description": "Output resolution of the short edge in pixels: 1440 for 1440p/2K or 1080 for 1080p/Full HD. The long edge follows the source aspect ratio, so portrait videos stay portrait. Default: 1440, or 1080 when the source
|
|
832
|
+
"description": "Output resolution of the short edge in pixels: 2160 for 2160p/4K (relayvsr only), 1440 for 1440p/2K, or 1080 for 1080p/Full HD. The long edge follows the source aspect ratio, so portrait videos stay portrait. Default: 1440, or 1080 when the source is too small for 1440p. Set 1080 when the user asks for 1080p or Full HD, and 2160 with model relayvsr when the user asks for 4K or 2160p."
|
|
824
833
|
},
|
|
825
834
|
"detailPreference": {
|
|
826
835
|
"type": "string",
|
|
@@ -828,7 +837,7 @@ export const SOGNI_HOSTED_TOOLS_MANIFEST: unknown = {
|
|
|
828
837
|
"stable",
|
|
829
838
|
"sharper"
|
|
830
839
|
],
|
|
831
|
-
"description": "Detail preference. stable, the default, gives the most temporally stable result. sharper renders crisper fine detail with a little more risk of shimmer between frames. Set sharper only when the user asks for a sharper, crisper, or more detailed upscale; otherwise omit it. It does not change the price."
|
|
840
|
+
"description": "FlashVSR only. Detail preference. stable, the default, gives the most temporally stable result. sharper renders crisper fine detail with a little more risk of shimmer between frames. Set sharper only when the user asks for a sharper, crisper, or more detailed upscale; otherwise omit it. It does not change the price."
|
|
832
841
|
},
|
|
833
842
|
"processingSpeed": {
|
|
834
843
|
"type": "string",
|
|
@@ -836,13 +845,13 @@ export const SOGNI_HOSTED_TOOLS_MANIFEST: unknown = {
|
|
|
836
845
|
"stable",
|
|
837
846
|
"faster"
|
|
838
847
|
],
|
|
839
|
-
"description": "Processing speed. stable, the default, gives the most stable result. faster finishes sooner with slightly less stable detail. Set faster only when the user asks for a quicker upscale; otherwise omit it. It does not change the price."
|
|
848
|
+
"description": "FlashVSR only. Processing speed. stable, the default, gives the most stable result. faster finishes sooner with slightly less stable detail. Set faster only when the user asks for a quicker upscale; otherwise omit it. It does not change the price."
|
|
840
849
|
},
|
|
841
850
|
"seed": {
|
|
842
851
|
"type": "integer",
|
|
843
852
|
"minimum": -1,
|
|
844
853
|
"maximum": 4294967295,
|
|
845
|
-
"description": "Seed for the upscaler's fine texture, 0 through 4294967295; different seeds give slightly different fine detail. Omit it to keep the repeatable default of 0. Use -1 for a random seed. Set it only when the user gives a seed or asks for a different or random variation of an upscale."
|
|
854
|
+
"description": "FlashVSR only. Seed for the upscaler's fine texture, 0 through 4294967295; different seeds give slightly different fine detail. Omit it to keep the repeatable default of 0. Use -1 for a random seed. Set it only when the user gives a seed or asks for a different or random variation of an upscale."
|
|
846
855
|
}
|
|
847
856
|
},
|
|
848
857
|
"required": []
|
|
@@ -444,12 +444,40 @@ export function validateAndNormalizeHostedToolArguments(
|
|
|
444
444
|
warnings: []
|
|
445
445
|
};
|
|
446
446
|
|
|
447
|
+
// upscale_video had no model argument in 5.83.0. A model value other than flashvsr
|
|
448
|
+
// or relayvsr is validated as it was then: kept, or stripped with the same warning.
|
|
449
|
+
let toolSchema = schema;
|
|
450
|
+
if (
|
|
451
|
+
toolName === 'upscale_video' &&
|
|
452
|
+
Object.prototype.hasOwnProperty.call(args, 'model') &&
|
|
453
|
+
args.model !== 'flashvsr' &&
|
|
454
|
+
args.model !== 'relayvsr' &&
|
|
455
|
+
schema.properties?.model
|
|
456
|
+
) {
|
|
457
|
+
const properties = { ...schema.properties };
|
|
458
|
+
delete properties.model;
|
|
459
|
+
toolSchema = { ...schema, properties };
|
|
460
|
+
}
|
|
461
|
+
|
|
447
462
|
const normalizedSchema: HostedToolSchema = {
|
|
448
|
-
...
|
|
449
|
-
type:
|
|
463
|
+
...toolSchema,
|
|
464
|
+
type: toolSchema.type ?? 'object'
|
|
450
465
|
};
|
|
451
466
|
const cleaned = validateValueAgainstSchema('', args, normalizedSchema, context).value;
|
|
452
467
|
|
|
468
|
+
// upscale_video: 2160 is RelayVSR only. FlashVSR (model omitted or anything but
|
|
469
|
+
// relayvsr) keeps the 1080/1440 refusal it had before the enum gained 2160.
|
|
470
|
+
if (toolName === 'upscale_video' && isRecord(cleaned)) {
|
|
471
|
+
const flashVsr2160Error = `Argument "targetResolution" must be one of ${formatEnum([1080, 1440])}`;
|
|
472
|
+
if (
|
|
473
|
+
cleaned.model !== 'relayvsr' &&
|
|
474
|
+
cleaned.targetResolution === 2160 &&
|
|
475
|
+
!context.errors.includes(flashVsr2160Error)
|
|
476
|
+
) {
|
|
477
|
+
context.errors.push(flashVsr2160Error);
|
|
478
|
+
}
|
|
479
|
+
}
|
|
480
|
+
|
|
453
481
|
return {
|
|
454
482
|
ok: context.errors.length === 0,
|
|
455
483
|
errors: context.errors,
|
package/src/Chat/tools.ts
CHANGED
|
@@ -84,7 +84,7 @@ export const refineResultTool: ToolDefinition = getHostedTool('refine_result');
|
|
|
84
84
|
export const changeAngleTool: ToolDefinition = getHostedTool('change_angle');
|
|
85
85
|
export const animatePhotoTool: ToolDefinition = getHostedTool('animate_photo');
|
|
86
86
|
|
|
87
|
-
// Promptless upscalers (RTX VSR image enlargement, FlashVSR
|
|
87
|
+
// Promptless upscalers (RTX VSR image enlargement, FlashVSR and RelayVSR video).
|
|
88
88
|
export const upscaleImageTool: ToolDefinition = getHostedTool('upscale_image');
|
|
89
89
|
export const upscaleVideoTool: ToolDefinition = getHostedTool('upscale_video');
|
|
90
90
|
|
|
@@ -34,6 +34,7 @@ import {
|
|
|
34
34
|
getVideoWorkflowType,
|
|
35
35
|
getVideoAssetRequirements,
|
|
36
36
|
isVideoModel,
|
|
37
|
+
isRelayVsrModel,
|
|
37
38
|
isVideoUpscaleModel,
|
|
38
39
|
isVideoSegmentationModel,
|
|
39
40
|
calculateVideoFrames,
|
|
@@ -48,6 +49,7 @@ import {
|
|
|
48
49
|
isMinimaxH3TurboModel,
|
|
49
50
|
isMinimaxH3BalancedModel,
|
|
50
51
|
isMinimaxH3FastH3ReferenceModel,
|
|
52
|
+
isMinimaxH3PrismModel,
|
|
51
53
|
isMinimaxH3ReferenceModel,
|
|
52
54
|
isMinimaxH3AudioGuideModel,
|
|
53
55
|
isMinimaxH3KeyframeModel,
|
|
@@ -700,19 +702,22 @@ function validateMinimaxH3Params(params: VideoProjectParams): void {
|
|
|
700
702
|
invalid('MiniMax H3 fps is fixed at 24. Omit fps or set it to 24.');
|
|
701
703
|
}
|
|
702
704
|
// FastH3 Reference ends in `_turbo` but is distilled to 8 steps, so it is
|
|
703
|
-
// neither a 4-step Turbo id nor a Balanced one.
|
|
705
|
+
// neither a 4-step Turbo id nor a Balanced one. Prism's recipe is 8 steps too.
|
|
704
706
|
const isFastH3Reference = isMinimaxH3FastH3ReferenceModel(params.modelId);
|
|
707
|
+
const isPrism = isMinimaxH3PrismModel(params.modelId);
|
|
705
708
|
const isTurbo = isMinimaxH3TurboModel(params.modelId);
|
|
706
709
|
const isBalanced = isMinimaxH3BalancedModel(params.modelId);
|
|
707
|
-
const expectedSteps = isFastH3Reference ? 8 : isTurbo ? 4 : isBalanced ? 8 : 20;
|
|
710
|
+
const expectedSteps = isFastH3Reference || isPrism ? 8 : isTurbo ? 4 : isBalanced ? 8 : 20;
|
|
708
711
|
if (params.steps !== undefined && params.steps !== expectedSteps) {
|
|
709
712
|
const tier = isFastH3Reference
|
|
710
713
|
? ' FastH3 Reference'
|
|
711
|
-
:
|
|
712
|
-
? '
|
|
713
|
-
:
|
|
714
|
-
? '
|
|
715
|
-
:
|
|
714
|
+
: isPrism
|
|
715
|
+
? ' Prism'
|
|
716
|
+
: isTurbo
|
|
717
|
+
? ' Turbo'
|
|
718
|
+
: isBalanced
|
|
719
|
+
? ' Balanced'
|
|
720
|
+
: '';
|
|
716
721
|
invalid(`MiniMax H3${tier} steps are fixed at ${expectedSteps}.`);
|
|
717
722
|
}
|
|
718
723
|
if (params.guidance !== undefined && params.guidance !== 1) {
|
|
@@ -1688,7 +1693,7 @@ const VIDEO_UPSCALE_TIMING_ERROR =
|
|
|
1688
1693
|
'Omit the source timing, or supply the source video’s exact frame count and frame rate.';
|
|
1689
1694
|
|
|
1690
1695
|
/**
|
|
1691
|
-
*
|
|
1696
|
+
* Upscale source timing is optional. The server probes the uploaded video and
|
|
1692
1697
|
* adopts its exact frame count and frame rate when they are omitted; values a
|
|
1693
1698
|
* caller does send must describe the source and are checked against it.
|
|
1694
1699
|
*
|
|
@@ -1829,27 +1834,48 @@ function applyVideoParams(
|
|
|
1829
1834
|
if (params.sam3VideoPrompt !== undefined)
|
|
1830
1835
|
throw new Error('sam3VideoPrompt requires the SAM 3 video selection model.');
|
|
1831
1836
|
if (isVideoUpscaleModel(params.modelId)) {
|
|
1832
|
-
|
|
1833
|
-
|
|
1834
|
-
|
|
1835
|
-
|
|
1836
|
-
throw new Error('
|
|
1837
|
-
|
|
1838
|
-
|
|
1839
|
-
|
|
1840
|
-
|
|
1841
|
-
|
|
1842
|
-
|
|
1843
|
-
|
|
1837
|
+
const relayVsr = isRelayVsrModel(params.modelId);
|
|
1838
|
+
const upscalerName = relayVsr ? 'RelayVSR' : 'FlashVSR';
|
|
1839
|
+
if (relayVsr) {
|
|
1840
|
+
if (params.detailPreference != null) throw new Error('RelayVSR has no detailPreference.');
|
|
1841
|
+
if (params.processingSpeed != null) throw new Error('RelayVSR has no processingSpeed.');
|
|
1842
|
+
if (params.seed != null) throw new Error('RelayVSR has no seed.');
|
|
1843
|
+
} else {
|
|
1844
|
+
if (
|
|
1845
|
+
params.detailPreference != null &&
|
|
1846
|
+
!['stable', 'sharper'].includes(params.detailPreference)
|
|
1847
|
+
) {
|
|
1848
|
+
throw new Error('FlashVSR detailPreference must be stable or sharper.');
|
|
1849
|
+
}
|
|
1850
|
+
if (
|
|
1851
|
+
params.processingSpeed != null &&
|
|
1852
|
+
!['stable', 'faster'].includes(params.processingSpeed)
|
|
1853
|
+
) {
|
|
1854
|
+
throw new Error('FlashVSR processingSpeed must be stable or faster.');
|
|
1855
|
+
}
|
|
1856
|
+
const seed = params.seed ?? 0;
|
|
1857
|
+
if (!Number.isInteger(seed) || seed < -1 || seed > 4294967295) {
|
|
1858
|
+
throw new Error(
|
|
1859
|
+
'FlashVSR seed must be -1 (random) or an integer from 0 through 4294967295.'
|
|
1860
|
+
);
|
|
1861
|
+
}
|
|
1844
1862
|
}
|
|
1845
1863
|
const resolution =
|
|
1846
1864
|
params.upscaleResolution ?? Math.min(Number(params.width), Number(params.height));
|
|
1847
|
-
if (
|
|
1865
|
+
if (relayVsr) {
|
|
1866
|
+
if (![1080, 1440, 2160].includes(resolution))
|
|
1867
|
+
throw new Error('Choose 1080p, 1440p or 2160p for RelayVSR video upscaling.');
|
|
1868
|
+
// The source is unknown here, so only a size the caller sends can be checked.
|
|
1869
|
+
if (Math.max(Number(params.width) || 0, Number(params.height) || 0) > 3840)
|
|
1870
|
+
throw new Error('RelayVSR output can be at most 3840 pixels on its long edge.');
|
|
1871
|
+
} else if (![1080, 1440].includes(resolution)) {
|
|
1848
1872
|
throw new Error('Choose 1080p or 1440p for video upscaling.');
|
|
1849
|
-
|
|
1873
|
+
}
|
|
1874
|
+
if (!params.referenceVideo)
|
|
1875
|
+
throw new Error(`${upscalerName} requires an uploaded referenceVideo.`);
|
|
1850
1876
|
validateVideoUpscaleTiming(params);
|
|
1851
1877
|
if (params.positivePrompt?.trim() || params.negativePrompt?.trim())
|
|
1852
|
-
throw new Error(
|
|
1878
|
+
throw new Error(`${upscalerName} is promptless.`);
|
|
1853
1879
|
if (
|
|
1854
1880
|
params.teacacheThreshold != null ||
|
|
1855
1881
|
params.trimEndFrame ||
|
|
@@ -2016,7 +2042,7 @@ function applyVideoParams(
|
|
|
2016
2042
|
: isSeedanceModel(params.modelId)
|
|
2017
2043
|
? 4
|
|
2018
2044
|
: 1;
|
|
2019
|
-
//
|
|
2045
|
+
// Video upscaling has no client-side maximum: the server's admission check owns
|
|
2020
2046
|
// the longest source it accepts and refuses a longer one itself.
|
|
2021
2047
|
const duration = isVideoUpscaleModel(params.modelId)
|
|
2022
2048
|
? validateNumber(params.duration, { min: minDuration, propertyName: 'Video duration' })
|
|
@@ -2107,8 +2133,10 @@ function applyVideoParams(
|
|
|
2107
2133
|
params.upscaleResolution ?? Math.min(Number(params.width), Number(params.height));
|
|
2108
2134
|
keyFrame.steps = 1;
|
|
2109
2135
|
keyFrame.seed = params.seed ?? 0;
|
|
2110
|
-
|
|
2111
|
-
|
|
2136
|
+
if (!isRelayVsrModel(params.modelId)) {
|
|
2137
|
+
keyFrame.detailPreference = params.detailPreference ?? 'stable';
|
|
2138
|
+
keyFrame.processingSpeed = params.processingSpeed ?? 'stable';
|
|
2139
|
+
}
|
|
2112
2140
|
keyFrame.generateAudio = true;
|
|
2113
2141
|
keyFrame.interpolation = 'none';
|
|
2114
2142
|
}
|
|
@@ -372,15 +372,20 @@ export interface MinimaxH3Keyframe {
|
|
|
372
372
|
* two-stage ID (`minimax-h3-fastvideo-omni-int8_r2v_turbo_2stage`) takes the
|
|
373
373
|
* same request and delivers twice the canvas, with the same 384/544/768 px
|
|
374
374
|
* canvas choices as the other Two-Stage IDs.
|
|
375
|
+
* - Prism (`minimax-h3-prism-int8_r2v`) is an optional `r2v` model for mature
|
|
376
|
+
* themes with its recipe fixed at 8 Euler steps. It takes the same
|
|
377
|
+
* references, keyframes, catalog LoRAs, frame grid and canvas as the other
|
|
378
|
+
* one-stage `r2v` IDs, at the 768p class only (no two-stage ID). It is
|
|
379
|
+
* subject to server authorization and needs `disableNSFWFilter: true`.
|
|
375
380
|
* - The `i2v` model accepts `referenceImage`, `referenceImageEnd`, or both, and
|
|
376
381
|
* requires at least one of them. The `flf2v` model requires both.
|
|
377
382
|
*
|
|
378
383
|
* #### MiniMax H3 intermediate keyframes (every workflow except `t2v`)
|
|
379
384
|
* - `keyframes` pins up to 8 still images at chosen frames between the first
|
|
380
|
-
* and last frame.
|
|
385
|
+
* and last frame. 24 ids accept it (`isMinimaxH3KeyframeModel()`): `i2v` and
|
|
381
386
|
* `flf2v` on every tier (Standard, Balanced, LightX2V Turbo, FastH3 Turbo and
|
|
382
387
|
* FastH3 Two-Stage), the six FastH3 Sound to Video ids (`ia2v`, `flfa2v`,
|
|
383
|
-
* `a2v`, one- and two-stage) and the
|
|
388
|
+
* `a2v`, one- and two-stage) and the eight `r2v` ids. Text-to-video and
|
|
384
389
|
* every other model reject a non-empty list.
|
|
385
390
|
* - Each entry is `{ image, frameIndex }`. `frameIndex` is the 0-based pixel
|
|
386
391
|
* frame at 24 fps (`Math.round(seconds * 24)`): an integer from 1 to
|
|
@@ -446,8 +451,9 @@ export interface MinimaxH3Keyframe {
|
|
|
446
451
|
* - `minimax-h3-ref2va-fp8_r2v` (standard) and
|
|
447
452
|
* `minimax-h3-ref2va-fp8_r2v_turbo` (four-step Turbo), and
|
|
448
453
|
* `minimax-h3-ref2va-fp8_r2v_balanced` (eight-step Larry v4 Balanced) condition on labelled reference material
|
|
449
|
-
* rather than on frame anchors, as
|
|
450
|
-
* (`minimax-h3-fastvideo-omni-int8_r2v_turbo`)
|
|
454
|
+
* rather than on frame anchors, as do FastH3 Reference
|
|
455
|
+
* (`minimax-h3-fastvideo-omni-int8_r2v_turbo`) and Prism
|
|
456
|
+
* (`minimax-h3-prism-int8_r2v`). The checkpoint accepts up to 9 reference
|
|
451
457
|
* images, 3 reference videos (24fps, 2-15 seconds each), and 3 reference
|
|
452
458
|
* audio clips, with at most 12 reference files in total.
|
|
453
459
|
* - All of those ceilings apply, and at least one visual reference (image or
|
|
@@ -510,14 +516,15 @@ export interface VideoProjectParams extends BaseProjectParams {
|
|
|
510
516
|
/** Track one selected object across an uploaded video, preserving every source frame. */
|
|
511
517
|
sam3VideoPrompt?: Sam3VideoPrompt;
|
|
512
518
|
/**
|
|
513
|
-
* FlashVSR delivery resolution on the shorter edge
|
|
514
|
-
* With
|
|
515
|
-
* server adopts the verified source's values. Values that are sent
|
|
519
|
+
* FlashVSR or RelayVSR delivery resolution on the shorter edge; 2160 is RelayVSR only. Source
|
|
520
|
+
* timing and audio are preserved. With an upscale `referenceVideo`, `frames`, `fps`, `width` and
|
|
521
|
+
* `height` may be omitted: the server adopts the verified source's values. Values that are sent
|
|
522
|
+
* must match the source.
|
|
516
523
|
*/
|
|
517
|
-
upscaleResolution?: 1080 | 1440;
|
|
518
|
-
/** FlashVSR detail preference. Defaults to stable (More Stable). */
|
|
524
|
+
upscaleResolution?: 1080 | 1440 | 2160;
|
|
525
|
+
/** FlashVSR detail preference. Defaults to stable (More Stable). RelayVSR refuses it. */
|
|
519
526
|
detailPreference?: 'stable' | 'sharper';
|
|
520
|
-
/** FlashVSR processing speed. Defaults to stable (More Stable). */
|
|
527
|
+
/** FlashVSR processing speed. Defaults to stable (More Stable). RelayVSR refuses it. */
|
|
521
528
|
processingSpeed?: 'stable' | 'faster';
|
|
522
529
|
/**
|
|
523
530
|
* Number of frames to generate. Most requests pass `duration` instead and let
|
|
@@ -525,7 +532,7 @@ export interface VideoProjectParams extends BaseProjectParams {
|
|
|
525
532
|
* it). Pass `frames` when positions inside the clip must be exact, as with
|
|
526
533
|
* MiniMax H3 `keyframes`, whose `frameIndex` values must fit the count; H3
|
|
527
534
|
* takes `124 + n*17` (124, 141, 158, ... 362). When both are passed,
|
|
528
|
-
* `duration` wins, except on
|
|
535
|
+
* `duration` wins, except on video upscales.
|
|
529
536
|
*/
|
|
530
537
|
frames?: number;
|
|
531
538
|
/**
|
|
@@ -697,7 +704,7 @@ export interface VideoProjectParams extends BaseProjectParams {
|
|
|
697
704
|
/**
|
|
698
705
|
* MiniMax H3 intermediate keyframes: up to 8 still images, each pinned at a
|
|
699
706
|
* chosen frame between the first and last frame. Accepted by every MiniMax H3
|
|
700
|
-
* workflow except text-to-video,
|
|
707
|
+
* workflow except text-to-video, 24 ids (`isMinimaxH3KeyframeModel()`):
|
|
701
708
|
* image-to-video (`i2v`), first/last-frame (`flf2v`), Sound to Video (`ia2v`,
|
|
702
709
|
* `flfa2v`, `a2v`) and Reference to Video (`r2v`). Any other model rejects a
|
|
703
710
|
* non-empty list, and an empty list is the same as omitting the field.
|
|
@@ -1599,7 +1606,7 @@ export interface VideoEstimateRequest {
|
|
|
1599
1606
|
includeMonthlyFairUse?: boolean;
|
|
1600
1607
|
/** Include Priority mode's 1.5× usage in the daily-capacity estimate. */
|
|
1601
1608
|
priorityMode?: boolean;
|
|
1602
|
-
/**
|
|
1609
|
+
/** Video upscale source geometry for an estimate. The server verifies the uploaded file at admission. */
|
|
1603
1610
|
sourceWidth?: number;
|
|
1604
1611
|
sourceHeight?: number;
|
|
1605
1612
|
tokenType: TokenType;
|
|
@@ -71,7 +71,8 @@ const MINIMAX_H3_VIDEO_MODEL_IDS = new Set([
|
|
|
71
71
|
'minimax-h3-ref2va-fp8_r2v_2stage',
|
|
72
72
|
'minimax-h3-ref2va-fp8_r2v_balanced_2stage',
|
|
73
73
|
'minimax-h3-fastvideo-omni-int8_r2v_turbo',
|
|
74
|
-
'minimax-h3-fastvideo-omni-int8_r2v_turbo_2stage'
|
|
74
|
+
'minimax-h3-fastvideo-omni-int8_r2v_turbo_2stage',
|
|
75
|
+
'minimax-h3-prism-int8_r2v'
|
|
75
76
|
]);
|
|
76
77
|
|
|
77
78
|
export function getEnhacementStrength(strength: EnhancementStrength): number {
|
|
@@ -111,8 +112,18 @@ export function isVideoSegmentationModel(modelId: string): boolean {
|
|
|
111
112
|
/** Standalone, promptless enhancement of a finished video. */
|
|
112
113
|
export const FLASHVSR_VIDEO_UPSCALE_MODEL_ID = 'flashvsr_v1.1_tiny_long_bf16';
|
|
113
114
|
|
|
115
|
+
/** RelayVSR: promptless upscaling at a fixed 2x or 4x, up to 2160p. */
|
|
116
|
+
export const RELAYVSR_X2_VIDEO_UPSCALE_MODEL_ID = 'relayvsr_x2_lightx2v_int8_convrot';
|
|
117
|
+
export const RELAYVSR_X4_VIDEO_UPSCALE_MODEL_ID = 'relayvsr_x4_int8_convrot';
|
|
118
|
+
|
|
119
|
+
export function isRelayVsrModel(modelId: string): boolean {
|
|
120
|
+
return (
|
|
121
|
+
modelId === RELAYVSR_X2_VIDEO_UPSCALE_MODEL_ID || modelId === RELAYVSR_X4_VIDEO_UPSCALE_MODEL_ID
|
|
122
|
+
);
|
|
123
|
+
}
|
|
124
|
+
|
|
114
125
|
export function isVideoUpscaleModel(modelId: string): boolean {
|
|
115
|
-
return modelId === FLASHVSR_VIDEO_UPSCALE_MODEL_ID;
|
|
126
|
+
return modelId === FLASHVSR_VIDEO_UPSCALE_MODEL_ID || isRelayVsrModel(modelId);
|
|
116
127
|
}
|
|
117
128
|
|
|
118
129
|
/**
|
|
@@ -439,13 +450,20 @@ export function isWan3EnhancedModel(modelId: string): boolean {
|
|
|
439
450
|
* `minimax-h3-fastvideo-omni-int8_r2v_turbo_2stage` takes the same request
|
|
440
451
|
* and delivers the clip at exactly twice the canvas, like the Ref2VA
|
|
441
452
|
* two-stage ids.
|
|
453
|
+
* - Prism: `minimax-h3-prism-int8_r2v`, an optional reference-to-video model
|
|
454
|
+
* for mature themes (a community merge of two Ref2VA finetunes) with its
|
|
455
|
+
* recipe fixed at 8 Euler steps (see `isMinimaxH3PrismModel`). It takes the
|
|
456
|
+
* same references, keyframes, catalog LoRAs and limits as the other one-stage
|
|
457
|
+
* r2v ids at the 768p class only: it has no two-stage id. Server
|
|
458
|
+
* authorization applies, and the Sensitive Content Filter must be off
|
|
459
|
+
* (`disableNSFWFilter: true`).
|
|
442
460
|
*
|
|
443
461
|
* All H3 paths share fixed 24fps, guidance 1, and the `124 + n*17` frame grid.
|
|
444
462
|
* Every path except the FastH3 audio guide generates 32kHz stereo audio
|
|
445
463
|
* jointly; audio-guide output carries the uploaded audio instead. Standard H3 uses 20 steps;
|
|
446
464
|
* Balanced uses qualified fixed 8-step acceleration: LightX2V for FL2VA and
|
|
447
465
|
* Larry v4 for Ref2VA; each Turbo family uses its own 4-step distillation LoRA;
|
|
448
|
-
* FastH3 Reference is distilled to a fixed 8 steps.
|
|
466
|
+
* FastH3 Reference is distilled to a fixed 8 steps, and so is Prism.
|
|
449
467
|
*/
|
|
450
468
|
export function isMinimaxH3Model(modelId: string): boolean {
|
|
451
469
|
return MINIMAX_H3_VIDEO_MODEL_IDS.has(modelId);
|
|
@@ -501,6 +519,15 @@ export function isMinimaxH3AudioGuideModel(modelId: string): boolean {
|
|
|
501
519
|
return /^minimax-h3-fastvideo-int8_(?:ia2v|flfa2v|a2v)_turbo(?:_2stage)?$/.test(modelId);
|
|
502
520
|
}
|
|
503
521
|
|
|
522
|
+
/**
|
|
523
|
+
* Check if a model ID is MiniMax H3 Prism (`minimax-h3-prism-int8_r2v`): a
|
|
524
|
+
* one-stage reference-to-video id (`isMinimaxH3ReferenceModel`) whose steps are
|
|
525
|
+
* fixed at 8. It is not a Balanced id: it carries no Larry v4 adapter.
|
|
526
|
+
*/
|
|
527
|
+
export function isMinimaxH3PrismModel(modelId: string): boolean {
|
|
528
|
+
return modelId === 'minimax-h3-prism-int8_r2v';
|
|
529
|
+
}
|
|
530
|
+
|
|
504
531
|
/**
|
|
505
532
|
* Check if a model ID is one of the 8-step MiniMax H3 Balanced workflows.
|
|
506
533
|
* FL2VA covers t2v/i2v/flf2v; Ref2VA uses its matching Larry v4 adapter for
|
|
@@ -517,8 +544,9 @@ export function isMinimaxH3BalancedModel(modelId: string): boolean {
|
|
|
517
544
|
/**
|
|
518
545
|
* Check if a model ID is the MiniMax H3 Ref2VA multi-reference workflow
|
|
519
546
|
* (`minimax-h3-ref2va-fp8_r2v`, `..._r2v_turbo`, `..._r2v_balanced`, the
|
|
520
|
-
* two-stage `..._r2v_2stage` and `..._r2v_balanced_2stage`,
|
|
521
|
-
* Reference `minimax-h3-fastvideo-omni-int8_r2v_turbo` and its `_2stage` id
|
|
547
|
+
* two-stage `..._r2v_2stage` and `..._r2v_balanced_2stage`, FastH3
|
|
548
|
+
* Reference `minimax-h3-fastvideo-omni-int8_r2v_turbo` and its `_2stage` id,
|
|
549
|
+
* or Prism `minimax-h3-prism-int8_r2v`).
|
|
522
550
|
*
|
|
523
551
|
* This is the only MiniMax H3 workflow that conditions on more than two input
|
|
524
552
|
* files, and the only video workflow of any family that carries reference
|
|
@@ -541,13 +569,13 @@ export const MINIMAX_H3_MAX_KEYFRAMES = 8;
|
|
|
541
569
|
* the worker pins (ComfyUI `MiniMaxH3AddGuide`) at chosen frames between the
|
|
542
570
|
* first and last frame.
|
|
543
571
|
*
|
|
544
|
-
* Every MiniMax H3 workflow except text-to-video accepts them,
|
|
572
|
+
* Every MiniMax H3 workflow except text-to-video accepts them, 24 ids in all:
|
|
545
573
|
* - image-to-video (`i2v`) and first/last-frame (`flf2v`) on every tier
|
|
546
574
|
* (Standard, Balanced, LightX2V Turbo, FastH3 Turbo and FastH3 Two-Stage);
|
|
547
575
|
* - the FastH3 Sound to Video audio guide (`ia2v`, `flfa2v`, `a2v`, one-stage
|
|
548
576
|
* and two-stage; `isMinimaxH3AudioGuideModel`);
|
|
549
|
-
* - Reference to Video (`r2v`: Standard, Turbo, Balanced, FastH3 Reference
|
|
550
|
-
* and
|
|
577
|
+
* - Reference to Video (`r2v`: Standard, Turbo, Balanced, FastH3 Reference,
|
|
578
|
+
* Prism and the three two-stage ids; `isMinimaxH3ReferenceModel`).
|
|
551
579
|
*/
|
|
552
580
|
export function isMinimaxH3KeyframeModel(modelId: string): boolean {
|
|
553
581
|
if (!isMinimaxH3Model(modelId)) return false;
|
package/src/index.ts
CHANGED
|
@@ -34,6 +34,8 @@ import {
|
|
|
34
34
|
MINIMAX_H3_MAX_KEYFRAMES,
|
|
35
35
|
PIXAL3D_IMAGE_TO_3D_MODEL_ID,
|
|
36
36
|
PIXAL3D_MULTIVIEW_IMAGE_TO_3D_MODEL_ID,
|
|
37
|
+
RELAYVSR_X2_VIDEO_UPSCALE_MODEL_ID,
|
|
38
|
+
RELAYVSR_X4_VIDEO_UPSCALE_MODEL_ID,
|
|
37
39
|
PIXAL3D_ORBIT_VIEW_SLOTS,
|
|
38
40
|
SAM3_IMAGE_SEGMENT_MODEL_ID,
|
|
39
41
|
SAM3_VIDEO_SEGMENT_MODEL_ID,
|
|
@@ -46,6 +48,7 @@ import {
|
|
|
46
48
|
isModelArtifactModel,
|
|
47
49
|
isPixal3dModel,
|
|
48
50
|
isPixal3dMultiViewModel,
|
|
51
|
+
isRelayVsrModel,
|
|
49
52
|
isSegmentationModel,
|
|
50
53
|
isVideoModel,
|
|
51
54
|
isVideoUpscaleModel,
|
|
@@ -278,6 +281,7 @@ import StatsApi from './Stats/index.js';
|
|
|
278
281
|
import ReplayApi from './Replay/index.js';
|
|
279
282
|
import WorldsApi from './Worlds/index.js';
|
|
280
283
|
import AnnouncementsApi from './Announcements/index.js';
|
|
284
|
+
import ChangelogApi from './Changelog/index.js';
|
|
281
285
|
import MediaEncryptionApi from './MediaEncryption/index.js';
|
|
282
286
|
import { MediaEncryptionError } from './lib/mediaEncryption/index.js';
|
|
283
287
|
import {
|
|
@@ -561,6 +565,8 @@ export {
|
|
|
561
565
|
PIXAL3D_ORBIT_VIEW_SLOTS,
|
|
562
566
|
SAM3_IMAGE_SEGMENT_MODEL_ID,
|
|
563
567
|
SAM3_VIDEO_SEGMENT_MODEL_ID,
|
|
568
|
+
RELAYVSR_X2_VIDEO_UPSCALE_MODEL_ID,
|
|
569
|
+
RELAYVSR_X4_VIDEO_UPSCALE_MODEL_ID,
|
|
564
570
|
SEEDREAM_5_PRO_UNCENSORED_MODEL_ID,
|
|
565
571
|
SEEDREAM_5_5_UNCENSORED_MODEL_ID,
|
|
566
572
|
calculateVideoFrames,
|
|
@@ -572,6 +578,7 @@ export {
|
|
|
572
578
|
isModelArtifactModel,
|
|
573
579
|
isPixal3dModel,
|
|
574
580
|
isPixal3dMultiViewModel,
|
|
581
|
+
isRelayVsrModel,
|
|
575
582
|
isSegmentationModel,
|
|
576
583
|
isSeedreamModel,
|
|
577
584
|
isVideoModel,
|
|
@@ -779,6 +786,13 @@ export class SogniClient {
|
|
|
779
786
|
* announcement per account so it stays dismissed on every device.
|
|
780
787
|
*/
|
|
781
788
|
announcements: AnnouncementsApi;
|
|
789
|
+
/**
|
|
790
|
+
* The public Sogni changelog (`/v1/changelog`): what changed for artists,
|
|
791
|
+
* developers and worker operators. `changelog.list(options)` reads one page
|
|
792
|
+
* of entries with filters and a cursor, `changelog.get(id)` one entry, and
|
|
793
|
+
* `changelog.products()` the products entries can name. No sign-in needed.
|
|
794
|
+
*/
|
|
795
|
+
changelog: ChangelogApi;
|
|
782
796
|
/**
|
|
783
797
|
* Optional media encryption: read encrypted results, video and
|
|
784
798
|
* audio. Create the client with `mediaEncryption: 'auto'` to encrypt every
|
|
@@ -804,6 +818,7 @@ export class SogniClient {
|
|
|
804
818
|
this.worlds = new WorldsApi(config);
|
|
805
819
|
this.replay = new ReplayApi(config);
|
|
806
820
|
this.announcements = new AnnouncementsApi(config);
|
|
821
|
+
this.changelog = new ChangelogApi(config);
|
|
807
822
|
|
|
808
823
|
this.apiClient = config.client;
|
|
809
824
|
}
|
|
@@ -965,6 +980,15 @@ export type {
|
|
|
965
980
|
WorldBuildWaiting,
|
|
966
981
|
WorldBuildWaitingReason
|
|
967
982
|
} from './Worlds/types.js';
|
|
983
|
+
export type {
|
|
984
|
+
ChangelogAudience,
|
|
985
|
+
ChangelogEntry,
|
|
986
|
+
ChangelogEntryType,
|
|
987
|
+
ChangelogLink,
|
|
988
|
+
ChangelogProduct,
|
|
989
|
+
ListChangelogOptions,
|
|
990
|
+
ListChangelogResult
|
|
991
|
+
} from './Changelog/types.js';
|
|
968
992
|
|
|
969
993
|
export {
|
|
970
994
|
getGptImageGifLayouts,
|
package/src/lib/validation.ts
CHANGED
package/src/version.ts
CHANGED