| 1 | import { z } from 'zod' |
| 2 | |
| 3 | // ========== Volcengine Track Structure Schema (matches volcengine.interface.ts) ========== |
| 4 | |
| 5 | const targetTimeSchema = z |
| 6 | .array(z.number()) |
| 7 | .length(2) |
| 8 | .transform(value => value as [number, number]) |
| 9 | |
| 10 | // TransformFilter - 2D Transform |
| 11 | export const transformFilterSchema = z.object({ |
| 12 | Type: z.literal('transform').describe('Filter type'), |
| 13 | PosX: z.number().describe('Element top-left X coordinate (pixels). Full-screen video MUST use 0, NOT canvas center!'), |
| 14 | PosY: z.number().describe('Element top-left Y coordinate (pixels). Full-screen video MUST use 0, NOT canvas center!'), |
| 15 | Width: z.number().describe('Element width (pixels)'), |
| 16 | Height: z.number().describe('Element height (pixels)'), |
| 17 | Rotation: z.number().optional().describe('Rotation angle [-360, 360], clockwise positive, rotates around element center'), |
| 18 | FlipX: z.boolean().optional().describe('Horizontal mirror flip'), |
| 19 | FlipY: z.boolean().optional().describe('Vertical mirror flip'), |
| 20 | Alpha: z.number().optional().describe('Opacity [0,1], 0 is transparent'), |
| 21 | }) |
| 22 | |
| 23 | // LutFilter - Color Filter |
| 24 | export const lutFilterSchema = z.object({ |
| 25 | Type: z.literal('lut_filter').describe('Filter type'), |
| 26 | TargetTime: targetTimeSchema.describe('Filter time range in element [startMs, endMs]'), |
| 27 | Source: z.string().describe('Filter ID'), |
| 28 | Intensity: z.number().optional().describe('Filter intensity [0,1]'), |
| 29 | }) |
| 30 | |
| 31 | // EffectFilter - Visual Effect |
| 32 | export const effectFilterSchema = z.object({ |
| 33 | Type: z.literal('effect_filter').describe('Filter type'), |
| 34 | TargetTime: targetTimeSchema.describe('Filter time range in element [startMs, endMs]'), |
| 35 | Source: z.string().describe('Effect ID'), |
| 36 | }) |
| 37 | |
| 38 | // TransitionFilter - Transition |
| 39 | export const transitionFilterSchema = z.object({ |
| 40 | Type: z.literal('transition').describe('Filter type'), |
| 41 | Source: z.string().describe('Transition ID'), |
| 42 | Duration: z.number().describe('Transition duration (milliseconds)'), |
| 43 | }) |
| 44 | |
| 45 | // CropFilter - Area Crop |
| 46 | export const cropFilterSchema = z.object({ |
| 47 | Type: z.literal('crop').describe('Filter type'), |
| 48 | PosX: z.number().describe('Crop area top-left X coordinate (pixels), NOT center point'), |
| 49 | PosY: z.number().describe('Crop area top-left Y coordinate (pixels), NOT center point'), |
| 50 | Width: z.number().describe('Crop area width (pixels)'), |
| 51 | Height: z.number().describe('Crop area height (pixels)'), |
| 52 | }) |
| 53 | |
| 54 | // TrimFilter - Time Trim |
| 55 | export const trimFilterSchema = z.object({ |
| 56 | Type: z.literal('trim').describe('Filter type'), |
| 57 | StartTime: z.number().describe('Trim start time (milliseconds)'), |
| 58 | EndTime: z.number().describe('Trim end time (milliseconds)'), |
| 59 | }) |
| 60 | |
| 61 | // SpeedFilter - Playback Speed |
| 62 | export const speedFilterSchema = z.object({ |
| 63 | Type: z.literal('speed').describe('Filter type'), |
| 64 | Speed: z.number().describe('Playback speed [0.1, 4]'), |
| 65 | }) |
| 66 | |
| 67 | // VideoAnimationsFilter - Video Animation |
| 68 | export const videoAnimationsFilterSchema = z.object({ |
| 69 | Type: z.literal('video_animation').describe('Filter type'), |
| 70 | AnimRes: z.string().describe('Video animation ID'), |
| 71 | AnimStartTime: z.number().describe('Animation start time (milliseconds)'), |
| 72 | AnimEndTime: z.number().describe('Animation end time (milliseconds)'), |
| 73 | AnimLoop: z.boolean().optional().describe('Whether to loop'), |
| 74 | AnimLoopDuration: z.number().optional().describe('Loop animation duration (milliseconds)'), |
| 75 | }) |
| 76 | |
| 77 | // TextAnimationsFilter - Text Animation |
| 78 | export const textAnimationsFilterSchema = z.object({ |
| 79 | Type: z.literal('text_animation').describe('Filter type'), |
| 80 | AnimInRes: z.string().optional().describe('Enter animation ID'), |
| 81 | AnimInDuration: z.number().optional().describe('Enter animation duration (milliseconds)'), |
| 82 | AnimOutRes: z.string().optional().describe('Exit animation ID'), |
| 83 | AnimOutDuration: z.number().optional().describe('Exit animation duration (milliseconds)'), |
| 84 | AnimLoopRes: z.string().optional().describe('Loop animation ID'), |
| 85 | AnimLoopDuration: z.number().optional().describe('Loop animation duration (milliseconds)'), |
| 86 | }) |
| 87 | |
| 88 | // AudioFadeFilter - Audio Fade |
| 89 | export const audioFadeFilterSchema = z.object({ |
| 90 | Type: z.literal('a_fade').describe('Filter type'), |
| 91 | FadeIn: z.number().optional().describe('Fade in time (milliseconds)'), |
| 92 | FadeOut: z.number().optional().describe('Fade out time (milliseconds)'), |
| 93 | }) |
| 94 | |
| 95 | // AudioVolumeFilter - Volume Control |
| 96 | export const audioVolumeFilterSchema = z.object({ |
| 97 | Type: z.literal('a_volume').describe('Filter type'), |
| 98 | Volume: z.number().describe('Volume, 0 is mute, 1 is default'), |
| 99 | }) |
| 100 | |
| 101 | // EqFilter - Image Equalizer |
| 102 | export const eqFilterSchema = z.object({ |
| 103 | Type: z.literal('equalizer').describe('Filter type'), |
| 104 | TargetTime: targetTimeSchema.describe('Filter time range in element [startMs, endMs]'), |
| 105 | Brightness: z.number().optional().describe('Brightness [0, 1]'), |
| 106 | Saturation: z.number().optional().describe('Saturation [0, 1]'), |
| 107 | Tone: z.number().optional().describe('Tone [0, 1]'), |
| 108 | }) |
| 109 | |
| 110 | // DelogoFilter - Local Blur |
| 111 | export const delogoFilterSchema = z.object({ |
| 112 | Type: z.literal('delogo').describe('Filter type'), |
| 113 | TargetTime: targetTimeSchema.describe('Filter time range in element [startMs, endMs]'), |
| 114 | DelogoType: z.string().optional().describe('Delogo type, default gaussian'), |
| 115 | PosX: z.number().describe('Local blur area top-left X coordinate (pixels), NOT center point'), |
| 116 | PosY: z.number().describe('Local blur area top-left Y coordinate (pixels), NOT center point'), |
| 117 | Width: z.number().describe('Local blur area width (pixels)'), |
| 118 | Height: z.number().describe('Local blur area height (pixels)'), |
| 119 | Sigma: z.number().optional().describe('Blur intensity [10,70]'), |
| 120 | Radius: z.number().optional().describe('Blur radius [10,70]'), |
| 121 | }) |
| 122 | |
| 123 | // GaussianFilter - Gaussian Blur |
| 124 | export const gaussianFilterSchema = z.object({ |
| 125 | Type: z.literal('gaussian').describe('Filter type'), |
| 126 | TargetTime: targetTimeSchema.describe('Filter time range in element [startMs, endMs]'), |
| 127 | Radius: z.number().describe('Gaussian blur radius [10,70]'), |
| 128 | Sigma: z.number().describe('Gaussian blur intensity [10,70]'), |
| 129 | }) |
| 130 | |
| 131 | // Filter Union Schema |
| 132 | export const filterSchema = z.union([ |
| 133 | transformFilterSchema, |
| 134 | lutFilterSchema, |
| 135 | effectFilterSchema, |
| 136 | transitionFilterSchema, |
| 137 | cropFilterSchema, |
| 138 | trimFilterSchema, |
| 139 | speedFilterSchema, |
| 140 | videoAnimationsFilterSchema, |
| 141 | textAnimationsFilterSchema, |
| 142 | audioFadeFilterSchema, |
| 143 | audioVolumeFilterSchema, |
| 144 | eqFilterSchema, |
| 145 | delogoFilterSchema, |
| 146 | gaussianFilterSchema, |
| 147 | ]).describe('Filter effect') |
| 148 | |
| 149 | // TrackElement Schema |
| 150 | export const trackElementSchema = z.object({ |
| 151 | Type: z.enum(['video', 'audio', 'image', 'text', 'subtitle', 'effect', 'gaussian', 'text_roll']).describe('Element type'), |
| 152 | Source: z.string().optional().describe('Resource source: vid://xxx, mid://xxx, tos://xxx or URL'), |
| 153 | TargetTime: targetTimeSchema.describe('Track time range [startMs, endMs], in milliseconds'), |
| 154 | // text/subtitle specific fields |
| 155 | Text: z.string().optional().describe('Text content'), |
| 156 | TextRes: z.string().optional().describe('Fancy text ID'), |
| 157 | FontType: z.string().optional().describe('Font ID'), |
| 158 | FontSize: z.number().optional().describe('Font size (pixels)'), |
| 159 | FontColor: z.string().optional().describe('Font color (RGBA)'), |
| 160 | BackgroundColor: z.string().optional().describe('Font background color (RGBA)'), |
| 161 | BorderColor: z.string().optional().describe('Font border color (RGBA)'), |
| 162 | BorderWidth: z.number().optional().describe('Font border width'), |
| 163 | LineMaxWidth: z.number().optional().describe('Auto line wrap width [-1, 1]'), |
| 164 | Typesetting: z.number().optional().describe('Text direction: 0=horizontal, 1=vertical'), |
| 165 | AlignType: z.number().optional().describe('Text alignment'), |
| 166 | // Extra - Filter list |
| 167 | Extra: z.array(filterSchema).optional().describe('Extra effects list (Filters)'), |
| 168 | }).describe('Track element') |
| 169 | |
| 170 | // Canvas Schema |
| 171 | export const canvasSchema = z.object({ |
| 172 | Width: z.number().describe('Canvas width (pixels)'), |
| 173 | Height: z.number().describe('Canvas height (pixels)'), |
| 174 | BackgroundColor: z.string().optional().describe('Canvas background color (RGBA)'), |
| 175 | }) |
| 176 | |
| 177 | // Codec Schema |
| 178 | export const codecSchema = z.object({ |
| 179 | VideoCodec: z.string().optional().describe('Video codec: h264, h265, vp9'), |
| 180 | AudioCodec: z.string().optional().describe('Audio codec: aac'), |
| 181 | VideoBitRate: z.number().optional().describe('Video bitrate (bps)'), |
| 182 | AudioBitrate: z.number().optional().describe('Audio bitrate (kbps)'), |
| 183 | Crf: z.number().optional().describe('Constant quality factor [0,51]'), |
| 184 | Preset: z.string().optional().describe('Encoding speed preset'), |
| 185 | }) |
| 186 | |
| 187 | // Output Schema |
| 188 | export const outputSchema = z.object({ |
| 189 | Fps: z.number().optional().describe('Output frame rate'), |
| 190 | DisableVideo: z.boolean().optional().describe('Output audio only'), |
| 191 | DisableAudio: z.boolean().optional().describe('Output video only'), |
| 192 | Alpha: z.boolean().optional().describe('Include Alpha channel'), |
| 193 | Codec: codecSchema.optional().describe('Codec settings'), |
| 194 | }) |
| 195 | |
| 196 | // submitDirectEditTask Schema |
| 197 | export const submitDirectEditTaskSchema = z.object({ |
| 198 | Canvas: canvasSchema.optional().describe('Canvas configuration (optional — auto-detected from primary video source if omitted)'), |
| 199 | Output: outputSchema.optional().describe('Output configuration'), |
| 200 | Track: z.array(z.array(trackElementSchema)).describe('Track array, outer array is render layer (higher index = higher layer), inner array is elements on same track'), |
| 201 | }) |
| 202 | |
| 203 | export const getVideoEditTaskStatusSchema = z.object({ |
| 204 | taskId: z.string().describe('Task ID (returned by submitDirectEditTask)'), |
| 205 | }) |
| 206 | |
| 207 | export enum VideoEditToolName { |
| 208 | SubmitDirectEditTask = 'submitDirectEditTask', |
| 209 | GetVideoEditTaskStatus = 'getVideoEditTaskStatus', |
| 210 | } |
| 211 |