tt/ttstd-infinite-canvas
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(*): 质量检查与版本发布 - 修复格式门禁和创作灵感兼容导出并发布 v1.6.2

Browse Source
This commit is contained in:
ddcat666
2026-10-08 20:43:15 +08:00
parent 04689257d2
commit 9174aae013
86 changed files with 22712 additions and 635 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
web/src/lib/canvas/image-operation-plan.ts
View File

@@ -36,9 +36,7 @@ export function resolveUpscaleSize(width: number, height: number, requested: num
assertPixelSize({ width, height });
if (!Number.isFinite(requested) || requested <= 0) throw new RangeError("目标尺寸无效");
const edge = Math.max(1, Math.min(MAX_UPSCALE_LONG_EDGE, Math.round(requested)));
return width >= height
? { width: edge, height: Math.max(1, Math.round((height / width) * edge)) }
: { width: Math.max(1, Math.round((width / height) * edge)), height: edge };
return width >= height ? { width: edge, height: Math.max(1, Math.round((height / width) * edge)) } : { width: Math.max(1, Math.round((width / height) * edge)), height: edge };
}
export function resolveCropPixels(size: PixelSize, selection?: ImageCropRect): ImageCropRect {

12
web/src/lib/image-utils.ts
View File

@@ -28,10 +28,14 @@ function decodeDataUrl(url: string): { bytes: Uint8Array<ArrayBuffer>; mimeType:
const encoded = new TextEncoder().encode(payload);
const output: number[] = [];
for (let cursor = 0; cursor < encoded.length; cursor++) {
if (encoded[cursor] !== 37) { output.push(encoded[cursor]); continue; }
if (encoded[cursor] !== 37) {
output.push(encoded[cursor]);
continue;
}
const hex = String.fromCharCode(...encoded.slice(cursor + 1, cursor + 3));
if (!/^[0-9a-f]{2}$/i.test(hex)) throw new Error("图片数据包含无效的转义字符");
output.push(parseInt(hex, 16)); cursor += 2;
output.push(parseInt(hex, 16));
cursor += 2;
}
bytes = Uint8Array.from(output);
}
@@ -47,7 +51,9 @@ export function getDataUrlByteSize(url: string): number {
return Math.floor(payload.replace(/=+$/, "").length * 0.75);
}
return decodeDataUrl(url).bytes.byteLength;
} catch { return 0; }
} catch {
return 0;
}
}
export async function readFileAsDataUrl(file: Blob): Promise<string> {

152
web/src/lib/inspirations/catalog.ts
View File

@@ -127,23 +127,141 @@ const curatedWorks: CreationInspiration[] = [
mode: "text",
prompt: "请帮我写一场发生在暖色灯光客厅里的母女对话戏：女儿准备离开家乡，母亲表面平静但一直在整理旧物，用动作和潜台词推进情绪。",
},
{ title: "森林里的最后一封信", description: "童话叙事 · 从角色愿望建立冲突", image: "/short-drama-styles/storybook-fantasy.jpg", mode: "text", sourceId: "chatgpt-prompts", credit: "Storyteller", prompt: "作为故事讲述者，为亲子观众写一个关于坚持的森林童话。主角是一只替动物送信的小狐狸，它必须在第一场雪之前送出一封没有地址的信。先给出角色的愿望与阻碍，再设计三个递进的事件，最后用一个可见的行动完成情绪转折。避免说教。" },
{ title: "末班车上的陌生人", description: "短剧编剧 · 让对白藏住真正目的", image: "/short-drama-styles/urban-live-action.jpg", mode: "text", sourceId: "chatgpt-prompts", credit: "Screenwriter", prompt: "作为编剧，设计一段发生在末班公交车上的三分钟短剧。两位陌生人都声称自己丢失了一本笔记，但只有一人说真话。先定义人物动机、场景限制和各自隐瞒的信息，再写动作、对白与反转结尾。全剧限制两个角色、一个场景，适合低成本实拍。" },
{ title: "未来城市的记忆修补师", description: "科幻大纲 · 世界规则与人物代价", image: "/short-drama-styles/future-tech.jpg", mode: "text", sourceId: "chatgpt-prompts", credit: "Novelist", prompt: "作为小说家，构思一部关于记忆修补师的未来都市短篇。城市居民能删除痛苦记忆，但主角发现每次删除都会遗失一种颜色。建立三条世界规则、主人公的核心缺陷、五个转折和付出代价的结局。先交付故事大纲，不直接铺写长篇。" },
{ title: "把告别写成一场雨", description: "诗性旁白 · 少解释，多感官细节", image: "/short-drama-styles/nature-healing.jpg", mode: "text", sourceId: "chatgpt-prompts", credit: "Poet", prompt: "作为诗人，为一段雨中离乡的影像写四组中文旁白，每组不超过四十字。分别使用气味、触感、声音与光线表达告别，不直接使用悲伤、不舍或想念等情绪词。用朴素、可朗读的语言，给每组留一个停顿。" },
{ title: "一瓶饮料的夏日故事", description: "广告策划 · 人群、主张与三个镜头", image: "/short-drama-styles/real-life.jpg", mode: "text", sourceId: "chatgpt-prompts", credit: "Advertiser", prompt: "作为广告策划，为面向年轻通勤者的无糖气泡饮料设计十五秒广告。定义目标人群、核心场景和一句传播主张，再给出三个镜头的画面、时长、音效与字幕。不要虚构营养或功效数据，用下班后一口清爽的体验讲故事。" },
{ title: "独立咖啡店的视觉语言", description: "品牌创意 · 让配色和语气讲同一件事", image: "/short-drama-styles/retro-hong-kong.jpg", mode: "text", sourceId: "chatgpt-prompts", credit: "Creative Branding Strategist", prompt: "作为品牌策略师，为社区里一家安静的独立咖啡店提出三种视觉方向。品牌重视慢节奏、邻里关系和手作。每种方向包含色彩、摄影氛围、文字语气、三条社交内容主题和差异化理由。不要模仿已有连锁品牌的标识。" },
{ title: "给剧本做一次镜头诊断", description: "创作复盘 · 从叙事到声音提出修改", image: "/short-drama-styles/black-white-noir.jpg", mode: "text", sourceId: "chatgpt-prompts", credit: "Film Critic", prompt: "请以电影评论者的视角审阅我接下来提供的剧本或分镜，从叙事、人物、构图、运镜和声音五个维度提出反馈。每条意见都引用我提供的片段，并给出一个能实际拍摄的修改方案。还没有收到材料时，先向我索取，不要假装看过影片。" },
{ title: "旧城街景的时代考据", description: "历史场景 · 区分史实、推测和虚构", image: "/short-drama-styles/period-live-action.jpg", mode: "text", sourceId: "chatgpt-prompts", credit: "Historian", prompt: "作为历史研究者，协助我设计民国早期城市街景的影视场景。先问清具体年份、地点和人物阶层，再列出服饰、交通、招牌、建筑和日常用语的考据清单。明确区分有来源的史实、合理推测与戏剧性虚构，不编造文献。" },
{ title: "月面独行", description: "远景到特写 · 太空探索的孤独感", image: "/short-drama-styles/space-opera.jpg", mode: "video", prompt: "月球背面的荒原，宇航员独自走向一座废弃天线。镜头从极远景缓慢推进，低角度阳光在月尘上拉出长影，步伐沉重，头盔反射星光，克制冷色，宽银幕电影构图，无字幕。" },
{ title: "手作星球早餐", description: "定格动画 · 柔软材质与小动作", image: "/short-drama-styles/clay-stop-motion.jpg", mode: "video", prompt: "黏土定格动画风格，一位小小宇航员在圆形厨房煎一颗星星形状的鸡蛋，鸡蛋轻轻跳起。手作指纹质感、暖奶油色、柔和侧光，固定中景，动作简短连续，可爱但不夸张。" },
{ title: "山水之间的一叶舟", description: "水墨留白 · 横向移镜与轻雾", image: "/short-drama-styles/ink-narrative.jpg", mode: "video", prompt: "中国水墨动画，一叶小舟缓缓穿过层叠远山，薄雾留白，舟上人抬手触碰落下的花瓣。镜头平稳横移，墨色有自然晕染，少量朱红点缀，画面安静，动作节奏舒缓。" },
{ title: "雨后的玻璃花房", description: "治愈场景 · 前景水滴与逆光", image: "/short-drama-styles/nature-healing.jpg", mode: "video", prompt: "清晨雨后的玻璃花房，镜头透过挂着水滴的叶片缓缓推向盛开的白花，阳光从侧后方进入，浅景深，真实植物纹理，空气中薄薄水汽，宁静自然，无人物和文字。" },
{ title: "都市少女角色转面", description: "角色概念 · 统一服装与形体比例", image: "/short-drama-styles/chinese-2d.jpg", mode: "image", prompt: "原创二维动画都市少女角色设定，短发、宽松工装夹克和帆布包，正面、侧面与背面三视图，服装细节和身高比例一致，干净线稿配平涂阴影，暖灰背景，不添加文字，不模仿已有动漫角色。" },
{ title: "梦境里的倒置房间", description: "超现实概念 · 重力反转与柔光", image: "/short-drama-styles/surreal-dream.jpg", mode: "image", prompt: "超现实电影场景概念图，一间漂浮在云海中的倒置卧室，床和书桌悬在天花板上，一扇门通向浅粉色天空。低饱和奶油与雾蓝色，柔和漫反射，精确透视，细腻材质，无人物，无文字。" },
{ title: "海边旅店的黄昏", description: "实拍场景 · 用道具留下人物痕迹", image: "/short-drama-styles/urban-live-action.jpg", mode: "image", prompt: "沿海小城旧旅店房间的电影置景概念图，窗外是黄昏海面，风吹动米白窗帘，桌上有两杯没有喝完的茶和一本翻开的书。暖橙与浅蓝对比，自然胶片颗粒，眼平中广角，没有人物和可读文字。" },
{ title: "鲜亮漫画封面", description: "平面构成 · 大色块与戏剧性透视", image: "/short-drama-styles/comic-pop.jpg", mode: "image", prompt: "原创漫画封面插画，一位骑单车的城市信使跃过雨后积水，戏剧性低角度透视，青蓝、珊瑚红与暖黄色大色块，半调网点，清晰黑色轮廓，背景简洁，为顶部标题留白但不生成文字。" },
{ title: "烘焙店的秘密帮手", description: "三维角色 · 柔和材质与亲切表情", image: "/short-drama-styles/three-d-cartoon.jpg", mode: "image", prompt: "原创三维动画角色概念图，一只穿着面粉围裙的小浣熊站在烘焙台前，鼻尖沾着面粉，手里举着刚烤好的面包。柔软毛发、温暖侧光、自然可爱的比例，橘棕色和奶油白配色，背景轻微虚化。" },
{
title: "森林里的最后一封信",
description: "童话叙事 · 从角色愿望建立冲突",
image: "/short-drama-styles/storybook-fantasy.jpg",
mode: "text",
sourceId: "chatgpt-prompts",
credit: "Storyteller",
prompt: "作为故事讲述者，为亲子观众写一个关于坚持的森林童话。主角是一只替动物送信的小狐狸，它必须在第一场雪之前送出一封没有地址的信。先给出角色的愿望与阻碍，再设计三个递进的事件，最后用一个可见的行动完成情绪转折。避免说教。",
},
{
title: "末班车上的陌生人",
description: "短剧编剧 · 让对白藏住真正目的",
image: "/short-drama-styles/urban-live-action.jpg",
mode: "text",
sourceId: "chatgpt-prompts",
credit: "Screenwriter",
prompt: "作为编剧，设计一段发生在末班公交车上的三分钟短剧。两位陌生人都声称自己丢失了一本笔记，但只有一人说真话。先定义人物动机、场景限制和各自隐瞒的信息，再写动作、对白与反转结尾。全剧限制两个角色、一个场景，适合低成本实拍。",
},
{
title: "未来城市的记忆修补师",
description: "科幻大纲 · 世界规则与人物代价",
image: "/short-drama-styles/future-tech.jpg",
mode: "text",
sourceId: "chatgpt-prompts",
credit: "Novelist",
prompt: "作为小说家，构思一部关于记忆修补师的未来都市短篇。城市居民能删除痛苦记忆，但主角发现每次删除都会遗失一种颜色。建立三条世界规则、主人公的核心缺陷、五个转折和付出代价的结局。先交付故事大纲，不直接铺写长篇。",
},
{
title: "把告别写成一场雨",
description: "诗性旁白 · 少解释，多感官细节",
image: "/short-drama-styles/nature-healing.jpg",
mode: "text",
sourceId: "chatgpt-prompts",
credit: "Poet",
prompt: "作为诗人，为一段雨中离乡的影像写四组中文旁白，每组不超过四十字。分别使用气味、触感、声音与光线表达告别，不直接使用悲伤、不舍或想念等情绪词。用朴素、可朗读的语言，给每组留一个停顿。",
},
{
title: "一瓶饮料的夏日故事",
description: "广告策划 · 人群、主张与三个镜头",
image: "/short-drama-styles/real-life.jpg",
mode: "text",
sourceId: "chatgpt-prompts",
credit: "Advertiser",
prompt: "作为广告策划，为面向年轻通勤者的无糖气泡饮料设计十五秒广告。定义目标人群、核心场景和一句传播主张，再给出三个镜头的画面、时长、音效与字幕。不要虚构营养或功效数据，用下班后一口清爽的体验讲故事。",
},
{
title: "独立咖啡店的视觉语言",
description: "品牌创意 · 让配色和语气讲同一件事",
image: "/short-drama-styles/retro-hong-kong.jpg",
mode: "text",
sourceId: "chatgpt-prompts",
credit: "Creative Branding Strategist",
prompt: "作为品牌策略师，为社区里一家安静的独立咖啡店提出三种视觉方向。品牌重视慢节奏、邻里关系和手作。每种方向包含色彩、摄影氛围、文字语气、三条社交内容主题和差异化理由。不要模仿已有连锁品牌的标识。",
},
{
title: "给剧本做一次镜头诊断",
description: "创作复盘 · 从叙事到声音提出修改",
image: "/short-drama-styles/black-white-noir.jpg",
mode: "text",
sourceId: "chatgpt-prompts",
credit: "Film Critic",
prompt: "请以电影评论者的视角审阅我接下来提供的剧本或分镜，从叙事、人物、构图、运镜和声音五个维度提出反馈。每条意见都引用我提供的片段，并给出一个能实际拍摄的修改方案。还没有收到材料时，先向我索取，不要假装看过影片。",
},
{
title: "旧城街景的时代考据",
description: "历史场景 · 区分史实、推测和虚构",
image: "/short-drama-styles/period-live-action.jpg",
mode: "text",
sourceId: "chatgpt-prompts",
credit: "Historian",
prompt: "作为历史研究者，协助我设计民国早期城市街景的影视场景。先问清具体年份、地点和人物阶层，再列出服饰、交通、招牌、建筑和日常用语的考据清单。明确区分有来源的史实、合理推测与戏剧性虚构，不编造文献。",
},
{
title: "月面独行",
description: "远景到特写 · 太空探索的孤独感",
image: "/short-drama-styles/space-opera.jpg",
mode: "video",
prompt: "月球背面的荒原，宇航员独自走向一座废弃天线。镜头从极远景缓慢推进，低角度阳光在月尘上拉出长影，步伐沉重，头盔反射星光，克制冷色，宽银幕电影构图，无字幕。",
},
{
title: "手作星球早餐",
description: "定格动画 · 柔软材质与小动作",
image: "/short-drama-styles/clay-stop-motion.jpg",
mode: "video",
prompt: "黏土定格动画风格，一位小小宇航员在圆形厨房煎一颗星星形状的鸡蛋，鸡蛋轻轻跳起。手作指纹质感、暖奶油色、柔和侧光，固定中景，动作简短连续，可爱但不夸张。",
},
{
title: "山水之间的一叶舟",
description: "水墨留白 · 横向移镜与轻雾",
image: "/short-drama-styles/ink-narrative.jpg",
mode: "video",
prompt: "中国水墨动画，一叶小舟缓缓穿过层叠远山，薄雾留白，舟上人抬手触碰落下的花瓣。镜头平稳横移，墨色有自然晕染，少量朱红点缀，画面安静，动作节奏舒缓。",
},
{
title: "雨后的玻璃花房",
description: "治愈场景 · 前景水滴与逆光",
image: "/short-drama-styles/nature-healing.jpg",
mode: "video",
prompt: "清晨雨后的玻璃花房，镜头透过挂着水滴的叶片缓缓推向盛开的白花，阳光从侧后方进入，浅景深，真实植物纹理，空气中薄薄水汽，宁静自然，无人物和文字。",
},
{
title: "都市少女角色转面",
description: "角色概念 · 统一服装与形体比例",
image: "/short-drama-styles/chinese-2d.jpg",
mode: "image",
prompt: "原创二维动画都市少女角色设定，短发、宽松工装夹克和帆布包，正面、侧面与背面三视图，服装细节和身高比例一致，干净线稿配平涂阴影，暖灰背景，不添加文字，不模仿已有动漫角色。",
},
{
title: "梦境里的倒置房间",
description: "超现实概念 · 重力反转与柔光",
image: "/short-drama-styles/surreal-dream.jpg",
mode: "image",
prompt: "超现实电影场景概念图，一间漂浮在云海中的倒置卧室，床和书桌悬在天花板上，一扇门通向浅粉色天空。低饱和奶油与雾蓝色，柔和漫反射，精确透视，细腻材质，无人物，无文字。",
},
{
title: "海边旅店的黄昏",
description: "实拍场景 · 用道具留下人物痕迹",
image: "/short-drama-styles/urban-live-action.jpg",
mode: "image",
prompt: "沿海小城旧旅店房间的电影置景概念图，窗外是黄昏海面，风吹动米白窗帘，桌上有两杯没有喝完的茶和一本翻开的书。暖橙与浅蓝对比，自然胶片颗粒，眼平中广角，没有人物和可读文字。",
},
{
title: "鲜亮漫画封面",
description: "平面构成 · 大色块与戏剧性透视",
image: "/short-drama-styles/comic-pop.jpg",
mode: "image",
prompt: "原创漫画封面插画，一位骑单车的城市信使跃过雨后积水，戏剧性低角度透视，青蓝、珊瑚红与暖黄色大色块，半调网点，清晰黑色轮廓，背景简洁，为顶部标题留白但不生成文字。",
},
{
title: "烘焙店的秘密帮手",
description: "三维角色 · 柔和材质与亲切表情",
image: "/short-drama-styles/three-d-cartoon.jpg",
mode: "image",
prompt: "原创三维动画角色概念图，一只穿着面粉围裙的小浣熊站在烘焙台前，鼻尖沾着面粉，手里举着刚烤好的面包。柔软毛发、温暖侧光、自然可爱的比例，橘棕色和奶油白配色，背景轻微虚化。",
},
];
/** 本项目编写的条目，灵感页把它单独作为「原创」分类。 */

62
web/src/lib/seedance-video.ts
View File

@@ -5,23 +5,44 @@ import type { ReferenceAudio, ReferenceVideo } from "@/types/media";
// Existing channel contract values; these are not a claim about every provider's current limits.
export const SEEDANCE_REFERENCE_LIMITS = { images: 9, videos: 3, audios: 3, imageMaxBytes: 30 * 1048576, videoMaxBytes: 50 * 1048576, audioMaxBytes: 15 * 1048576 };
export const seedanceResolutionOptions = [{ label: "480P", value: "480p" }, { label: "720P", value: "720p" }, { label: "1080P", value: "1080p" }] as const;
export const seedanceResolutionOptions = [
{ label: "480P", value: "480p" },
{ label: "720P", value: "720p" },
{ label: "1080P", value: "1080p" },
] as const;
export const seedanceRatioOptions = [
{ label: "横屏", value: "16:9" }, { label: "竖屏", value: "9:16" }, { label: "方形", value: "1:1" },
{ label: "标准横屏", value: "4:3" }, { label: "标准竖屏", value: "3:4" }, { label: "宽银幕", value: "21:9" }, { label: "自适应", value: "adaptive" },
{ label: "横屏", value: "16:9" },
{ label: "竖屏", value: "9:16" },
{ label: "方形", value: "1:1" },
{ label: "标准横屏", value: "4:3" },
{ label: "标准竖屏", value: "3:4" },
{ label: "宽银幕", value: "21:9" },
{ label: "自适应", value: "adaptive" },
] as const;
export const seedanceDurationOptions = VIDEO_DURATION_OPTIONS;
const rasterRows: Record<string, readonly string[]> = {
"16:9": ["864x496", "1280x720", "1920x1080"], "4:3": ["752x560", "1112x834", "1664x1248"],
"1:1": ["640x640", "960x960", "1440x1440"], "3:4": ["560x752", "834x1112", "1248x1664"],
"9:16": ["496x864", "720x1280", "1080x1920"], "21:9": ["992x432", "1470x630", "2206x946"],
"16:9": ["864x496", "1280x720", "1920x1080"],
"4:3": ["752x560", "1112x834", "1664x1248"],
"1:1": ["640x640", "960x960", "1440x1440"],
"3:4": ["560x752", "834x1112", "1248x1664"],
"9:16": ["496x864", "720x1280", "1080x1920"],
"21:9": ["992x432", "1470x630", "2206x946"],
};
const ratios = Object.keys(rasterRows).map((label) => { const [w, h] = label.split(":").map(Number); return { label, ratio: w / h }; });
const ratios = Object.keys(rasterRows).map((label) => {
const [w, h] = label.split(":").map(Number);
return { label, ratio: w / h };
});
export function isSeedanceVideoModel(model: string): boolean { return /seedance/i.test(model); }
export function isSeedanceFastModel(model: string): boolean { return isSeedanceVideoModel(model) && /fast/i.test(model); }
export function isArkPlanBaseUrl(baseUrl: string): boolean { return /\/api\/plan\/v3/i.test(baseUrl); }
export function isSeedanceVideoModel(model: string): boolean {
return /seedance/i.test(model);
}
export function isSeedanceFastModel(model: string): boolean {
return isSeedanceVideoModel(model) && /fast/i.test(model);
}
export function isArkPlanBaseUrl(baseUrl: string): boolean {
return /\/api\/plan\/v3/i.test(baseUrl);
}
export function isSeedanceVideoConfig(config: AiConfig | Pick<AiConfig, "model" | "videoModel" | "baseUrl">): boolean {
const request = "channels" in config ? resolveModelRequestConfig(config, config.model || config.videoModel) : config;
@@ -39,7 +60,9 @@ export function normalizeSeedanceResolution(value: string, model = ""): string {
const supported = isSeedanceFastModel(model) ? ["480p", "720p"] : ["480p", "720p", "1080p", "2160p"];
return supported.includes(requested) ? requested : "720p";
}
export function normalizeSeedanceDuration(value: string): number { return +normalizeVideoDuration(value); }
export function normalizeSeedanceDuration(value: string): number {
return +normalizeVideoDuration(value);
}
export function normalizeSeedanceRatio(value: string): string {
if (Object.hasOwn(rasterRows, value)) return value;
@@ -50,7 +73,10 @@ export function normalizeSeedanceRatio(value: string): string {
let distance = Infinity;
for (const entry of ratios) {
const next = Math.abs(entry.ratio - requested);
if (next < distance) { distance = next; nearest = entry; }
if (next < distance) {
distance = next;
nearest = entry;
}
}
return nearest.label;
}
@@ -61,11 +87,15 @@ export function seedancePixelLabel(resolution: string, ratio: string): string {
const column = ["480p", "720p", "1080p"].indexOf(normalizeSeedanceResolution(resolution));
return column < 0 ? "" : rasterRows[aspect]?.[column] || "";
}
export function boolConfig(value: string | undefined, fallback: boolean): boolean { return value === "true" || (value !== "false" && fallback); }
export function seedanceReferenceLabel(kind: "image" | "video" | "audio", index: number): string {
return `${({ image: "图片", video: "视频", audio: "音频" })[kind]}${index + 1}`;
export function boolConfig(value: string | undefined, fallback: boolean): boolean {
return value === "true" || (value !== "false" && fallback);
}
export function seedanceReferenceLabel(kind: "image" | "video" | "audio", index: number): string {
return `${{ image: "图片", video: "视频", audio: "音频" }[kind]}${index + 1}`;
}
export function buildSeedancePromptText(prompt: string, _images: ReferenceImage[], _videos: ReferenceVideo[], _audios: ReferenceAudio[]): string {
return prompt.trim();
}
export function buildSeedancePromptText(prompt: string, _images: ReferenceImage[], _videos: ReferenceVideo[], _audios: ReferenceAudio[]): string { return prompt.trim(); }
type VideoRule = { valid: (video: ReferenceVideo) => boolean; message: string };
const videoRules: VideoRule[] = [