feat: add video generation commands (SeeDance2.0)

README.md

@@ -1,6 +1,6 @@
 # ListenHub CLI
 
-Command-line interface for [ListenHub](https://listenhub.ai) — create podcasts, text-to-speech, explainer videos, slides, AI images, and music from your terminal.
+Command-line interface for [ListenHub](https://listenhub.ai) — create podcasts, text-to-speech, explainer videos, slides, AI images, music, and videos from your terminal.
 
 [中文文档](README.zh-CN.md)
 
@@ -76,6 +76,15 @@ listenhub tts create --text "Hello, world" --lang en
 | `listenhub image list`     | List AI images       |
 | `listenhub image get <id>` | Get image details    |
 
+### Video Generation
+
+| Command                    | Description                    |
+| -------------------------- | ------------------------------ |
+| `listenhub video create`   | Create a video generation task |
+| `listenhub video list`     | List video tasks               |
+| `listenhub video get <id>` | Get video task details         |
+| `listenhub video estimate` | Estimate credit cost           |
+
 ### Other
 
 | Command                             | Description             |
@@ -109,6 +118,9 @@ listenhub music cover --audio ./song.mp3
 # Local image for reference (jpg, png, webp, gif; max 10MB)
 listenhub image create --prompt "inspired by this" --reference ./photo.jpg
 
+# Local video for reference (mp4, mov; max 50MB)
+listenhub video create --prompt "same style" --reference-video ./clip.mp4 --input-video-duration 5
+
 # URLs are passed through directly
 listenhub music cover --audio https://example.com/song.mp3
 ```
@@ -158,6 +170,23 @@ listenhub image create \
   --aspect-ratio 16:9 --size 4K
 ```
 
+### Video generation
+
+```bash
+# Text-to-video
+listenhub video create --prompt "A cat playing piano in a jazz bar"
+
+# Image-to-video (first frame)
+listenhub video create --prompt "Camera slowly zooms out" --first-frame ./scene.png
+
+# With reference video
+listenhub video create --prompt "Same style dancing" \
+  --reference-video ./clip.mp4 --input-video-duration 8
+
+# Estimate credits
+listenhub video estimate --model doubao-seedance-2-pro --resolution 1080p --duration 10
+```
+
 ### JSON output for scripting
 
 ```bash

README.zh-CN.md

@@ -1,6 +1,6 @@
 # ListenHub CLI
 
-[ListenHub](https://listenhub.ai) 的命令行工具 — 在终端里创建播客、语音合成、讲解视频、幻灯片、AI 图片和音乐。
+[ListenHub](https://listenhub.ai) 的命令行工具 — 在终端里创建播客、语音合成、讲解视频、幻灯片、AI 图片、音乐和视频。
 
 [English](README.md)
 
@@ -76,6 +76,15 @@ listenhub tts create --text "你好世界" --lang zh
 | `listenhub image list`     | 列出图片     |
 | `listenhub image get <id>` | 查看图片详情 |
 
+### 视频生成
+
+| 命令                       | 说明             |
+| -------------------------- | ---------------- |
+| `listenhub video create`   | 创建视频生成任务 |
+| `listenhub video list`     | 列出视频任务     |
+| `listenhub video get <id>` | 查看视频任务详情 |
+| `listenhub video estimate` | 预估积分消耗     |
+
 ### 其他
 
 | 命令                                | 说明         |
@@ -100,7 +109,7 @@ listenhub tts create --text "你好世界" --lang zh
 
 ## 本地文件上传
 
-`music cover` 和 `image create` 支持引用本地文件。CLI 自动检测本地路径，校验格式和大小，上传到云存储后传给 API。
+`music cover`、`image create` 和 `video create` 支持引用本地文件。CLI 自动检测本地路径，校验格式和大小，上传到云存储后传给 API。
 
 ```bash
 # 本地音频文件用于翻唱（mp3, wav, flac, m4a, ogg, aac；最大 20MB）
@@ -109,6 +118,9 @@ listenhub music cover --audio ./song.mp3
 # 本地图片用于参考（jpg, png, webp, gif；最大 10MB）
 listenhub image create --prompt "以此为灵感" --reference ./photo.jpg
 
+# 本地视频用于参考（mp4, mov；最大 50MB）
+listenhub video create --prompt "同样风格" --reference-video ./clip.mp4 --input-video-duration 5
+
 # URL 直接透传
 listenhub music cover --audio https://example.com/song.mp3
 ```
@@ -158,6 +170,23 @@ listenhub image create \
   --aspect-ratio 16:9 --size 4K
 ```
 
+### 视频生成
+
+```bash
+# 文字生成视频
+listenhub video create --prompt "一只猫在爵士酒吧弹钢琴"
+
+# 图生视频（首帧）
+listenhub video create --prompt "镜头缓缓拉远" --first-frame ./scene.png
+
+# 带参考视频
+listenhub video create --prompt "相同风格的舞蹈" \
+  --reference-video ./clip.mp4 --input-video-duration 8
+
+# 预估积分
+listenhub video estimate --model doubao-seedance-2-pro --resolution 1080p --duration 10
+```
+
 ### 脚本中使用 JSON 输出
 
 ```bash

package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@marswave/listenhub-cli",
-	"version": "0.0.4",
+	"version": "0.0.5",
 	"description": "Command-line interface for ListenHub",
 	"license": "MIT",
 	"repository": "marswaveai/listenhub-cli",
@@ -25,7 +25,7 @@
 		"prepublishOnly": "pnpm run build"
 	},
 	"dependencies": {
-		"@marswave/listenhub-sdk": "^0.0.4",
+		"@marswave/listenhub-sdk": "^0.0.6",
 		"commander": "^14.0.3",
 		"open": "^10.0.0",
 		"ora": "^8.0.0"

source/_shared/mp4-duration.ts

@@ -0,0 +1,91 @@
+import {open} from 'node:fs/promises';
+
+export async function getMp4Duration(filePath: string): Promise<number> {
+	const file = await open(filePath, 'r');
+	try {
+		const moovOffset = await findAtom(file, 'moov', 0, await fileSize(file));
+		if (moovOffset === undefined) {
+			throw new Error(`Cannot read video duration: moov atom not found in ${filePath}`);
+		}
+
+		const moovHeader = await readAtomHeader(file, moovOffset);
+		const moovEnd = moovOffset + moovHeader.size;
+		const mvhdOffset = await findAtom(file, 'mvhd', moovOffset + 8, moovEnd);
+		if (mvhdOffset === undefined) {
+			throw new Error(`Cannot read video duration: mvhd atom not found in ${filePath}`);
+		}
+
+		const dataOffset = mvhdOffset + 8;
+
+		const versionBuf = Buffer.alloc(1);
+		await file.read(versionBuf, 0, 1, dataOffset);
+		const version = versionBuf[0]!;
+
+		let timescale: number;
+		let duration: bigint;
+
+		if (version === 0) {
+			const buf = Buffer.alloc(8);
+			await file.read(buf, 0, 8, dataOffset + 4 + 8);
+			timescale = buf.readUInt32BE(0);
+			duration = BigInt(buf.readUInt32BE(4));
+		} else if (version === 1) {
+			const buf = Buffer.alloc(12);
+			await file.read(buf, 0, 12, dataOffset + 4 + 16);
+			timescale = buf.readUInt32BE(0);
+			duration = buf.readBigUInt64BE(4);
+		} else {
+			throw new Error(`Cannot read video duration: unsupported mvhd version ${String(version)}`);
+		}
+
+		if (timescale === 0) {
+			throw new Error(`Cannot read video duration: timescale is 0`);
+		}
+
+		return Math.round(Number(duration) / timescale);
+	} finally {
+		await file.close();
+	}
+}
+
+interface AtomHeader {
+	size: number;
+	type: string;
+}
+
+async function readAtomHeader(
+	file: Awaited<ReturnType<typeof open>>,
+	offset: number,
+): Promise<AtomHeader> {
+	const buf = Buffer.alloc(8);
+	const {bytesRead} = await file.read(buf, 0, 8, offset);
+	if (bytesRead < 8) {
+		return {size: 0, type: ''};
+	}
+
+	const size = buf.readUInt32BE(0);
+	const type = buf.toString('ascii', 4, 8);
+	return {size, type};
+}
+
+async function findAtom(
+	file: Awaited<ReturnType<typeof open>>,
+	target: string,
+	start: number,
+	end: number,
+): Promise<number | undefined> {
+	let offset = start;
+	while (offset < end) {
+		const header = await readAtomHeader(file, offset); // eslint-disable-line no-await-in-loop
+		if (header.size === 0) break;
+		if (header.type === target) return offset;
+		offset += header.size;
+	}
+
+	return undefined;
+}
+
+async function fileSize(file: Awaited<ReturnType<typeof open>>): Promise<number> {
+	const stat = await file.stat();
+	return stat.size;
+}

source/_shared/polling.ts

@@ -4,6 +4,7 @@ import type {
 	ListenHubClient,
 	LyricsTaskDetail,
 	MusicTaskDetail,
+	VideoGenerationTaskDetail,
 } from '@marswave/listenhub-sdk';
 import ora from 'ora';
 import {CliTimeoutError} from './output.js';
@@ -127,6 +128,42 @@ export async function pollMusicTaskUntilDone(
 	throw new CliTimeoutError(`Timed out after ${timeoutS}s`);
 }
 
+export async function pollVideoTaskUntilDone(
+	client: ListenHubClient,
+	taskId: string,
+	options: {timeout?: number; json?: boolean},
+): Promise<VideoGenerationTaskDetail> {
+	const timeoutS = options.timeout ?? 1200;
+	const maxAttempts = Math.ceil(timeoutS / (pollIntervalMs / 1000));
+	const spinner = options.json
+		? undefined
+		: ora({text: `Generating video... (1/${maxAttempts})`}).start();
+
+	for (let i = 0; i < maxAttempts; i++) {
+		if (i > 0) {
+			await sleep(pollIntervalMs); // eslint-disable-line no-await-in-loop
+		}
+
+		const task = await client.getVideoGenerationTask(taskId); // eslint-disable-line no-await-in-loop
+		if (task.status === 'success') {
+			spinner?.succeed('Video created successfully');
+			return task;
+		}
+
+		if (task.status === 'failed') {
+			spinner?.fail('Video creation failed');
+			throw new Error('Video creation failed');
+		}
+
+		if (spinner) {
+			spinner.text = `Generating video... (${String(i + 2)}/${maxAttempts})`;
+		}
+	}
+
+	spinner?.fail('Timed out');
+	throw new CliTimeoutError(`Timed out after ${timeoutS}s`);
+}
+
 const lyricsIntervalMs = 5000;
 
 export async function pollLyricsTaskUntilDone(

source/_shared/upload.ts

@@ -2,19 +2,22 @@ import {access, readFile, stat} from 'node:fs/promises';
 import path from 'node:path';
 import type {ListenHubClient} from '@marswave/listenhub-sdk';
 
-type FileAcceptType = 'audio' | 'image';
+type FileAcceptType = 'audio' | 'image' | 'video';
 
 const audioExtensions = new Set(['.mp3', '.wav', '.flac', '.m4a', '.ogg', '.aac']);
 const imageExtensions = new Set(['.jpg', '.jpeg', '.png', '.webp', '.gif']);
+const videoExtensions = new Set(['.mp4', '.mov']);
 
 const maxSizeBytes: Record<FileAcceptType, number> = {
 	audio: 20 * 1024 * 1024,
 	image: 10 * 1024 * 1024,
+	video: 50 * 1024 * 1024,
 };
 
 const categoryForType: Record<FileAcceptType, string> = {
 	audio: 'episode',
 	image: 'banana',
+	video: 'episode',
 };
 
 const mimeTypes = new Map<string, string>([
@@ -29,16 +32,20 @@ const mimeTypes = new Map<string, string>([
 	['.png', 'image/png'],
 	['.webp', 'image/webp'],
 	['.gif', 'image/gif'],
+	['.mp4', 'video/mp4'],
+	['.mov', 'video/quicktime'],
 ]);
 
 function allowedExtensions(accept: FileAcceptType): Set<string> {
-	return accept === 'audio' ? audioExtensions : imageExtensions;
+	if (accept === 'audio') return audioExtensions;
+	if (accept === 'video') return videoExtensions;
+	return imageExtensions;
 }
 
 export async function resolveFileOrUrl(
 	client: ListenHubClient,
 	input: string,
-	options: {accept: FileAcceptType},
+	options: {accept: FileAcceptType; category?: string},
 ): Promise<string> {
 	const trimmed = input.trim();
 
@@ -77,7 +84,7 @@ export async function resolveFileOrUrl(
 	// Get presigned upload URL
 	const contentType = mimeTypes.get(ext)!;
 	const fileKey = path.basename(filePath);
-	const category = categoryForType[options.accept];
+	const category = options.category ?? categoryForType[options.accept];
 	const {presignedUrl, fileUrl} = await client.createFileUpload({
 		fileKey,
 		contentType,

source/cli.ts

@@ -10,6 +10,7 @@ import {register as registerPodcast} from './podcast/_cli.js';
 import {register as registerSlides} from './slides/_cli.js';
 import {register as registerSpeakers} from './speakers/_cli.js';
 import {register as registerTts} from './tts/_cli.js';
+import {register as registerVideo} from './video/_cli.js';
 
 const program = new Command();
 program.name('listenhub').description('ListenHub CLI').version('0.1.0');
@@ -23,6 +24,7 @@ registerImage(program);
 registerMusic(program);
 registerLyrics(program);
 registerSpeakers(program);
+registerVideo(program);
 registerCreation(program);
 
 program.parse();