← Back to Skills Marketplace
binkes

Chanjing One Click Video Creation

by BinKes · GitHub ↗ · v1.0.10 · MIT-0
cross-platform ✓ Security Clean
245
Downloads
0
Stars
1
Active Installs
11
Versions
Install in OpenClaw
/install chanjing-one-click-video-creation
Description
用户输入选题或工作流,自动生成完整短视频成片(文案、分镜、数字人口播与 AI 画面混剪);调用 Chanjing Open API 与同仓库子技能脚本。
README (SKILL.md)

一键式视频渲染器

功能说明

一键调用 Chanjing API 完成口播 TTS、数字人合成、文生视频与本地封装;集成 ffmpeg / ffprobe 做拼接、转码与轨对齐。编排与安全细则见 §3§8templates/成片命令§5

运行依赖

必须可用的二进制或等价封装(具体调用方式见 §5run_render.py):

  • ffmpeg:拼接、转码、封装
  • ffprobe:时长、分辨率、旋转等元数据(与数字人轨对齐)
  • chan-skill(或同仓库下直接 python 调用子技能脚本):驱动 chanjing-ttschanjing-video-composechanjing-ai-creation 等 CLI

环境变量与机器可读声明

  • 可选环境变量覆盖(均有默认或路径推断、非运行前必填):scripts/run_render.py 顶部及子 skill 源码;凭据路径manifest.yaml credentials
  • 合规 permissionsclientPermissionsagentPolicy、凭据模型:manifest.yaml

§3.2run_render.py 对齐;若与 manifest.yaml 冲突,以 manifest.yaml 为准

使用命令

  • ClawHub(slug 以注册表为准,常与技能包名一致):clawhub run chanjing-one-click-video-creation
  • 本仓库直连python scripts/run_render.py --input workflow.json --output-dir ./outputs/run1(在技能目录或配合 SKILLS_DIR / CHANJING_ONE_CLICK_VIDEO_SKILLS_ROOT / CHAN_SKILLS_DIR 使用)

速查

内容 位置
工作流、duration_secnull/合并、选题校验 §4.1
切段、奇偶镜、scenes[]scene_count/video_type首镜 voiceover ≤20 字(硬) storyboard_prompt.md 篇首「文本切段」;script_prompt.md 首镜口播;video_brief_plan.md
渲染技术、状态、partial/success、硬约束 render_rules.md §1–§4;§7§8
ref_prompt / 文生提示词 storyboard_prompt.md + history_storyboard_prompt.md§4.2 指针
请求体字段与默认 §6
run_render.py、子进程 CLI §5
安全、凭据、信任边界、策略 manifest.yaml + §3.1(§3.1 不重复 manifest 表格)
环境变量、二进制、副作用、落盘 §3.2

冲突:渲染实现以 render_rules.md 为准;ref_prompt 条文以 storyboard_prompt.md / history_storyboard_prompt.md 为准(§4.2 汇总指针)。run_render.py 只实现 §5 + render_rules.md,不增业务规则。执行:手工编排子 skill、仅 run_render、或混用。


1. 做什么

  1. 选题或全文 → video_plan、口播全文、分镜
  2. TTS:整段优先;超长按分镜少批合并(细则与字数见 render_rules.md §3·C.4
  3. 按镜切音频
  4. 数字人分镜chanjing-video-compose(音频驱动)
  5. AI 分镜ref_promptchanjing-ai-creation → 与镜内音频合成
  6. 封装:对齐公共数字人轨 → ffmpeg concat → 本地 mp4

2. 何时用 / 何时不用

适合 要成片;口播与画面混剪;用户明确要生成短视频
不适合 仅文案/标题;未要视频;只剪已有素材

3. 前置条件

  • 鉴权chanjing-credentials-guard;凭据路径与字段见 manifest.yaml§3.1;无凭证时子进程可 open_login_page.py
  • Plan/Script/分镜:本地 Agent 逻辑,无需外部 LLM API key(本 skill 必选路径不依赖外部 LLM)
  • 本机二进制与仓库布局§3.2ffmpeg / ffprobeSKILLS_DIR 等)
  • 数字人与音色用环境变量或仓库内缓存文件保存跨任务的「默认」audio_man / person_id / figure_type。每次任务在 workflow.json 根级显式填写;由 Agent 按 video_plan(如 video_type)、口播人设与选题语义,调用 list_voices.pylist_figures.py--sourcecommon / customised 等与本次任务一致)选型后写入;audio_man 宜与所选形象的 audio_man_id 一致。
  • 公共数字人选型(禁止「只取列表前几项」):须用 list_figures.py --source common --json 拉取候选(必要时增大 --page-size 或翻页,覆盖足够条目),在候选内逐项对比后再定稿:namefigures[].type(→ figure_type)、figures[].width/height(画幅与 D.1c 一致)、audio_man_idaudio_name(若有)与 video_plan/口播人设(性别、气质、行业、年龄感)是否匹配。默认偏好年轻、有活力的形象:名称或 audio_name 中含青年/少女/小哥哥/小姐姐/学生/元气/青春/年轻等正向信号时优先;仅当选题或用户明确要求成熟、权威、中老年等气质时,再选对应人设。定制源 customised 同样对比 namewidth/heightaudio_man_id 等,勿未经比较直接取页首。

3.1 安全、凭据与信任边界

环境变量与二进制以 manifest.yaml§3.2 为据。审阅时可对照 descriptionmanifest.yaml(含 agentPolicy)。

  • 能力与管道:步骤级说明见 §1run_render 职责与子进程见 §5(不在此重复链路)。
  • 主凭据 / 路径 / primaryEnv:见 manifest.yaml;路径与写回行为另见 §3.2 持久性表「凭据状态」及 CHANJING_OPENAPI_CREDENTIALS_DIR(兼容 CHANJING_CONFIG_DIR)。
  • 敏感与合规:勿回显完整密钥、勿将 credentials.json 提交版本库;权限建议 0700 / 0600(配置脚本尽量设置)。
  • 信任与出站行为:HTTPS、按返回 URL 拉取媒体、--output-dir 落盘等细节见 §3.2「典型副作用」与持久性表;须自行判断是否信任蝉镜主机与链接。
  • 浏览器:缺凭证时的 webbrowser.open / open_login_page.py§3.2 同表。
  • Agent 策略manifest.yamlagentPolicy(非 always、不改其它 skill)。

3.2 运行时契约(环境变量、二进制、副作用与落盘)

scripts/run_render.py 及同仓库子 skill 行为对齐;与篇首 YAML、manifest.yaml 一致。若与其它产品文档并列,以本仓库源码、manifest.yaml 与本文为准

环境变量(可选覆盖)

命名以仓库根 合规规则.md §3 为准。清单再逐项列举可选变量:不设亦可运行。凭据目录manifest.yaml credentialsdirectoryEnvdefaultPath)。run_render.pyopenapi_base_urlone_click_skills_repository_rootone_click_ref_prompt_max_charsone_click_ai_creation_model_code 及子进程环境即本技能可选覆盖与兼容旧名的实现位置。

说明:部分外部文档中的 FIRST_DIGITAL_HUMAN_MAX_CHARS 等变量,当前 run_render.py 未读取

外部二进制

二进制 必需性 用途
ffmpeg 跑一键成片 run_render.py 时必需 拼接、转码、封装音视频等。仅编排纯 API、不执行本渲染脚本时可不装。
ffprobe 同上 读取媒体分辨率、时长、旋转元数据等,用于与数字人轨对齐。

执行脚本时的典型副作用(按类)

类型 说明
出站 HTTPS 蝉镜 Open API(CHANJING_OPENAPI_BASE_URL / 兼容 CHANJING_API_BASE)、以及接口/CDN 返回的 video_url / 音频 URL 等素材拉取。
本地文件 run_render.py --output-dir 下常见:final_one_click.mp4workflow_result.jsonwork/(中间音频、分段视频、concat 列表等);具体以当次命令与 templates/render_rules.md 为准。
子进程 ffmpeg / ffproberun_render 通过 subprocess 调用同仓库下 skills/chanjing-ttschanjing-video-composechanjing-ai-creation 等目录中的 Python CLI。
浏览器 凭据缺失或引导登录时,鉴权链可能 webbrowser.open 或执行 chanjing-credentials-guardopen_login_page.py(与各 skill 的 _auth.py 行为一致)。

持久性变更范围与用户可控性

以下对本 skill 而言属预期内副作用;可通过路径与环境变量控制写入位置,而非隐式污染无关目录。

类别 写入什么 典型位置 用户如何控制
凭据状态 经配置写入的 app_id / secret_key、刷新后的 access_tokenexpire_in CHANJING_OPENAPI_CREDENTIALS_DIR/credentials.json(默认 ~/.chanjing/credentials.json;兼容 CHANJING_CONFIG_DIR 设置推荐名或旧名;或迁移/删除该文件;将秘钥提交版本库。
一键成片工件 final_one_click.mp4workflow_result.jsonwork/ run_render.py --output-dir 指定(常见为某次任务下的 outputs/\x3C任务名>/ 选用明确的 --output-dir;任务结束后按需保留或删除该目录。
其它下载类脚本(子 skill) 合成结果等到本地 各 skill 的 download_result.py 等:默认多在当前工作目录下 outputs/\x3C产品线>/,或 --output 绝对路径 在预期 cwd 下执行,或始终传 --output;详见对应 skill 的 SKILL.md
临时/过程文件 TTS 合并、切段、上传前缓存等 多在上述 output-dir 下的 work/ 或脚本约定子目录 随输出目录一并管理。

凭据文件:路径与读写语义见上表 凭据状态 行;首次配置见 skills/chanjing-credentials-guard/SKILL.md


4. 规则汇编

4.1 工作流编排

合并null = 不覆盖。顺序:默认铺底 → 非 null 覆盖 → 布尔/整数校正。字段默认见 §6;未在表中展开的缺省由 run_render.py(及子进程)按实现与环境变量读取(不含音色/数字人:audio_manperson_id/avatar_idfigure_type 仅来自 workflow.json,见 §3)。

duration_sec:策划参考,非 ffmpeg 上限。成片时长以 TTS+ffprobe 为准。scene_countvideo_brief_plan.md;切段与 AI 条数依实测与字幕轴(render_rules.md §3·C.5)。禁止为凑时长裁已定稿口播(除非用户要求)。

选题:去空白 \x3C5 字、占位串(如「你好」「test」)拒收;可扩写;严格模式模糊则失败。

步骤:1) Plan → video_brief_plan(败则全败;模板见 video_brief_plan.md)2) Script(hook / 首段与首镜对齐≤20 字硬上限,见 script_prompt.md)3) Storyboard:语义切分;storyboard_prompt.md首个分镜 voiceover 同上硬上限);非当代 history_storyboard_prompt.md;DH chanjing-video-compose,AI chanjing-ai-creation;TTS/多段 AI/mux render_rules.md §3§5 4) Render:render_rules.md §3(含 §3·C.6)、§4(表 4–6);ref_prompt 质检见 storyboard_prompt.md / history_storyboard_prompt.md§4.2);重试/partial render_rules.md §1 5) 成功:render_rules.md §1

仅渲染run_render.py + full_script + scenes[]顺序:Plan → Script → Storyboard → Render(各阶段用哪份模板见上列步骤)。


4.2 文生视频提示词(ref_prompt)— 指针

唯一条文真值(修订以模板为准,本文不重复 D.1–D.4 表文):

范围 模板
当代向、D.0 语境缺省与文明圈推断、D.1 长度、D.1aD.1b(易幻觉,全 skill 共用)、D.2 当代、手工 visual_prompt、D.3、D.4 当代装配与 7 要素 / 题材簇 / 单镜拼装 / 自检 templates/storyboard_prompt.md「文生视频提示词(当代向真值)」
D.2 非当代路由、历史流程层文明圈与国别自洽、占位符纪律、与 D.3/D.4 衔接说明 templates/history_storyboard_prompt.md
族裔、历史/非当代中式造型与出现人物时的英文短语 templates/visual_prompt_people_constraint.md(显式族裔锚定、历史 / 非当代节;兼 render_rules.md §4 表 4–6)

仍仅在此处索引:长音频多段 render_rules.md §3·C.6;字数上限 CHANJING_ONE_CLICK_VIDEO_REF_PROMPT_MAX_CHARS(兼容 AI_VIDEO_PROMPT_MAX_CHARS)。模板与 render_rules.md 实现冲突时以 render_rules.md 为准。


5. 自动化编排(run_render.py

依赖:鉴权;SKILLS_DIR / CHANJING_ONE_CLICK_VIDEO_SKILLS_ROOT / CHAN_SKILLS_DIR(§3);chanjing-tts / chanjing-video-compose / chanjing-ai-creation

职责:① TTS+audio_task_state;批合并与单批字数上限见 render_rules.md §3·C.4TTS_BATCH_MAX)② 切段(render_rules.md §3·C.5)③ 有 AI 镜时先完成首条数字人并 ffprobe(含 rotate)→ 再按映射提交文生 aspect_ratio/clarity(见 render_rules.md §3·C.6debug.ai_video_submit_params)④ 与其余 DH/AI 并行 poll ⑤ AI 轨对齐该参照 ffprobe ⑥ ffmpeg concat ⑦ 多段文生在 ref_prompt 后追加英文分层;总长由 CHANJING_ONE_CLICK_VIDEO_REF_PROMPT_MAX_CHARS(兼容 AI_VIDEO_PROMPT_MAX_CHARS)约束

不做:不产 plan/script/storyboard;不自动非当代/当代;不用 list_tasks.py 当代次(render_rules.md §4 表项 8

手工编排:仍须满足 render_rules.md §3、§4 与 §5;§3 细化(如 silencedetectminterpolate、参照轨码率、同套切段音频换形象、TTS 批间静音等)全部保留

输入 MVP

字段 必填 说明
full_script 与各镜 voiceoverscene_id 拼,norm 一致
scenes scene_idvoiceoveruse_avatar;AI 镜 ref_promptstoryboard_prompt.md / history_storyboard_prompt.md;§4.2);可选 subtitle
audio_man 宜与所选数字人形象的 audio_man_id 一致
person_id/avatar_id 条件 有 DH 镜必填
figure_type 与当次 list_figures.py 所选形象行的 figure_type 一致(公共多形态时必填)
subtitle_required 默认 false;为 true 时数字人镜烧录字幕(--subtitle show
speed/pitch 默认 1/1
ai_video_duration_sec 5 或 10,默认 10
model_code 默认 CHANJING_ONE_CLICK_VIDEO_CREATION_MODEL_CODE(兼容 AI_VIDEO_MODEL)或 Doubao-Seedance-1.0-pro;creation_type=4;不传 ref_img_url
max_retry_per_step 默认 1(§6)
python scripts/run_render.py --input workflow.json --output-dir ./outputs/run1

输出final_one_click.mp4workflow_result.jsonwork/


6. 输入(请求体)

norm:去 \r、首尾空白;空→空串;与 run_render.py 一致。口播:先 full_script,再 scriptcopy_textinput_scriptcontent 首个非空。无 topic:首句代选题(40 字内遇句末标点截,否则 24 字)。null/合并 §4.1

字段 必填 说明
topic 条件 无则见首句规则;建议 ≥5 字
industry/platform/style industry 空;platform/style:DEFAULT_*douyin/观点型口播
duration_sec DEFAULT_DURATION 或 60;策划参考
use_avatar 默认 true
avatar_id/voice_id 空;不得用环境变量兜底音色或数字人;须在 workflow.json 写明 audio_man/person_id(及有 DH 镜时的 figure_type),由 Agent 按当次任务调用 list_voices.pylist_figures.py(来源与 video_plan / 用户指定一致)对比 name、形态、画幅、audio_name 等后选型;禁止未比较即取列表最前几条;默认偏好年轻数字人(见 §3
subtitle_required 默认 false(数字人成片不烧录字幕;run_renderhide
cover_required 默认 true
strict_validation/allow_auto_expand_topic/max_retry_per_step true/false/1
full_script 默认空
script_title/script_hook/script_cta 默认空
script/… 见上文口播顺序

7. 输出 JSON

含义
status success / partial / failed
video_plan Plan
script_result title、hook、full_script、cta
storyboard_result.scenes[] scene_id、duration_sec、voiceover、subtitle、visual_prompt、use_avatar
render_result video_file、scene_video_urls、render_path、degrade_log
其它 error、debug…

渲染无降级:任一步失败即中断,不自动改为仅 DH 或仅 AI 成片。partial:未成 success(如 run_render 异常仍写 workflow_result.json);表示允许上述降级,storyboard_prompt.md·D.1b 类质检。成功 degrade_log=[];失败尽量保留已产出文案与分镜。


8. 硬性约束

表在 templates/render_rules.md §4;与 ref_prompt 交叉见 storyboard_prompt.md / history_storyboard_prompt.md§4.2 指针)。本节为锚点。


9. 限制

  • 本地 mp4;不上传
  • AI 单段常 5–10s;长口播多段
  • 成片时长=TTS 总轨;可与 duration_sec 不符
  • TTS:整轨优先、超长少批合并;单批上限与合并策略(含 TTS_BATCH_MAX)以 render_rules.md §3·C.4 为准
  • 文生失败可能为平台/模型;试增 max_retry_per_step、短 ref_prompt、拆镜;查 workflow_result.json
Usage Guidance
This skill appears to do what it says: orchestrate Chanjing TTS/video APIs, download media, and run local ffmpeg to produce a final mp4. Before installing or running it: 1) Verify you trust the Chanjing endpoints (API + CDN) because the skill will POST your credentials-derived token and download media URLs returned by the API. 2) Note credentials are persisted to credentials.json (default ~/.chanjing/credentials.json or directory set by CHANJING_OPENAPI_CREDENTIALS_DIR); follow the project guidance (0600/0700, don't commit to VCS). 3) Ensure ffmpeg/ffprobe and required sibling skills (chanjing-tts, chanjing-video-compose, chanjing-ai-creation or chan-skill) are present and on PATH. 4) If you need stricter network or filesystem controls, review manifest.yaml's allowed_hosts and filesystem roots and confirm whether res.chanjing.cc (CDN media) is acceptable for your deployment. 5) The templates include content-generation rules and stylistic biases (e.g., default preference for younger‑looking figures); review those if content neutrality is important. If you want extra assurance, inspect the sibling skills' _auth.py and the credentials guard flow (open_login_page.py) to see exactly how tokens are obtained and persisted.
Capability Analysis
Type: OpenClaw Skill Name: chanjing-one-click-video-creation Version: 1.0.10 The skill bundle is a legitimate video rendering automation tool designed to orchestrate TTS, digital human synthesis, and AI video generation via the Chanjing Open API. The core logic in `scripts/run_render.py` safely utilizes `subprocess` with argument lists to invoke `ffmpeg`, `ffprobe`, and sibling python scripts, and uses `urllib.request` to download media assets from documented API endpoints (open-api.chanjing.cc). The extensive markdown instructions in the `templates/` directory are strictly focused on prompt engineering for video quality and historical accuracy, containing no evidence of malicious prompt injection or unauthorized data access. Permissions defined in `manifest.yaml` are broad but align with the functional requirements of a media processing agent.
Capability Assessment
Purpose & Capability
Name/description (one‑click video creation using Chanjing APIs + ffmpeg) match the files and runtime behavior: run_render.py calls sibling Chanjing skills, polls TTS/video task state, downloads media, and uses ffmpeg/ffprobe to normalize and concat. The manifest and SKILL.md document the same dependencies and credentials model.
Instruction Scope
Runtime instructions and scripts perform network calls to Chanjing Open API, download returned media URLs, execute sibling skill CLIs via subprocess, call ffmpeg/ffprobe, and may open a browser for login. All of these are consistent with the described video-rendering pipeline; however the skill persists access_token on disk (credentials.json) and will read/write the credentials directory described in manifest.yaml — be aware of token persistence and where credentials are stored.
Install Mechanism
No install spec is present; the package is instruction + Python scripts that rely on existing binaries (ffmpeg/ffprobe) and sibling skills. This is lowest install risk and consistent with the declared approach (no arbitrary downloads/installers).
Credentials
The skill does not require unrelated cloud credentials; it uses a local credentials.json (app_id/secret_key/access_token) as expected for the Chanjing API. Manifest flags sensitive fields and opts to persist access_token on disk — expected but a privacy surface. Optional environment variables are documented but not required. One minor inconsistency: runtime examples/downloads use res.chanjing.cc (CDN) while manifest.allowed_hosts/documentedHosts list open-api.chanjing.cc and www.chanjing.cc — the skill intends to download media URLs returned by the API (manifest allows 'mayDownloadFromApiResponseUrls'), but you should confirm you trust the CDN domain used by the API.
Persistence & Privilege
always:false and agentPolicy modify flags are reasonable. The skill persists an access_token to credentials.json by design (manifest.persistAccessTokenOnDisk: true) and writes outputs under a user-controlled --output-dir; these are expected for a rendering pipeline and do not indicate silent privilege escalation. It does import and execute sibling skill code and invoke subprocesses (documented in manifest) — normal for this multi-skill orchestration.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install chanjing-one-click-video-creation
  3. After installation, invoke the skill by name or use /chanjing-one-click-video-creation
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.10
Version 1.0.10 - Added detailed intermediate and result output files including scene and concat states (e.g., `final_concat.txt`, `scene_times.json`, `tts_state.json`, `workflow_result.json`). - Updated `SKILL.md` to simplify and clarify environment variable descriptions and remove redundant tables. - Adjusted environment variable handling in documentation to specify only optional overlays, matching current script logic. - No changes to core features or inputs/outputs; all new files aid pipeline transparency and debugging.
v1.0.9
**Summary: Migrated to manifest.yaml as the primary machine-readable declaration, with updated environment variable conventions and improved documentation alignment.** - Added `manifest.yaml` for machine-readable metadata, describing environments, requirements, permissions, and policy. - Updated environment variable names to use new recommended forms (e.g., `CHANJING_OPENAPI_BASE_URL`) with legacy names still supported for compatibility. - Revised SKILL.md to reference and align with `manifest.yaml`, reducing duplication and clarifying precedence (manifest is authoritative). - Clarified credential file location, usage, and migration behavior in both documentation and environment variable references. - Improved quick reference tables and documentation for easier understanding and maintenance.
v1.0.8
Version 1.0.8 - Expanded and clarified skill metadata in SKILL.md to align with ClawHub/Openclaw requirements. - Added explicit fields for author, binaries, environment variables, category, and tags. - Improved descriptions of workflow, binary/environment dependencies, and command usage. - Enhanced quick reference tables and summary of side effects/persistence for better usability. - No changes to core code or functional behavior; documentation update only.
v1.0.7
### chanjing-one-click-video-creation 1.0.7 - Overhauled and simplified the SKILL.md frontmatter to focus on machine-readable YAML for registry and introspection. - Moved formerly verbose registration, credential, and agent policy explanations to structured YAML (`metadata` section); corresponding narrative in §3 is condensed and cross-referenced. - Clarified credential fields, file location, and omission of `primaryEnv` via machine-readable and prose locations. - Aggregated run-time requirements, side effects, and file outputs into a tabular "contract" for transparency. - No functional code changes—SKILL.md now prioritizes readability, automation, and explicit policy.
v1.0.6
**Summary: Major metadata simplification and credential documentation update.** - Simplified the metadata section in SKILL.md, removing `credentialModel` and `agentPolicy`; only required binaries and homepage now declared. - Updated description and documentation to clarify credential handling (file-based dual key, not OpenClaw primaryEnv). - Rewrote and condensed credential and environment variable usage notes for clarity and easier compliance review. - Removed redundant and overly detailed registry metadata and repeated information from user-facing documentation. - Clarified workflow and requirement references to match actual implementation and reduce confusion.
v1.0.5
**Expanded registry metadata for credential discovery and sensitive data review:** - Added machine-readable credentialModel to metadata, describing credential file location, sensitive fields, and do-not-commit guidance. - Clarified in description and documentation that credentials (app_id, secret_key, access_token) and token expiry are always stored in a local JSON file (not via environment variable). - Highlighted that install/onboarding requires explicit acceptance of disk/network sensitive surface: credential file persistence and download of media by API-returned URLs. - Confirmed primaryEnv is intentionally omitted, aligning metadata and behavior. - No changes to functional workflow or required binaries (ffmpeg/ffprobe).
v1.0.4
**Summary:** Moved TTS, digital human, and AI video logic to require external sub-skills; clarified credentials, dependencies, and registry details. - Migrated TTS, digital human, and AI video implementations out of internal `clients/`, now relying on sub-skills (no longer bundled within this skill). - Simplified credentials handling: now always read/write `app_id` and `secret_key` (and token) from `credentials.json`; no longer requires OpenClaw `primaryEnv`. - Updated documentation to clarify the credential model, required/optional environment variables (like `CHANJING_API_BASE` and `CHANJING_CONFIG_DIR`), and binary dependencies (`ffmpeg`, `ffprobe`). - Added explicit responses to registry and security review concerns (purpose alignment, instruction scope, credential persistence, trust boundaries). - Removed obsolete internal client modules (`clients/ai_creation_client.py`, etc.) and updated references in README and SKILL.md accordingly.
v1.0.3
Version 1.0.3 – No file changes detected. - No code or documentation files were modified in this release. - Skill behavior, workflow, and interface remain unchanged from the previous version.
v1.0.2
chanjing-one-click-video-creation v1.0.2 - Skill now encapsulates TTS, digital human video synthesis, and AI-generated visuals (文生视频) capabilities internally; no external skill dependencies required for core rendering. - Updated workflow to explicitly require digital human/audio model selection per task, with strict candidate comparison based on task context. - New documentation structure: quick reference, process rules, required input/output fields, and detailed orchestration steps. - Rendering pipeline (run_render.py) now fully self-contained and stops on any failure; no longer supports partial or degraded output modes. - Default preferences for digital human selection emphasize youthful, energetic personas unless user/task requires otherwise.
v1.0.1
changelog for v1.0.1: - Documentation updates in SKILL.md and README.md to clarify workflow, input requirements, digital human selection, and rendering constraints. - Refined input and output specification tables, emphasizing field requirements and default behaviors. - Improved description and references of rendering rules and prompt templates. - No business logic changes; updates focus on clarity and user guidance.
v1.0.0
chanjing-one-click-video-creation v1.0.0 - Initial release: Generate complete short videos from a user topic, including script, storyboard, digital human narration, and AI-generated visuals. - Supports one-click video creation and topic-based video generation workflows. - Integrates TTS, digital human (DH), and AI scene synthesis; manages audio/video segmentation and composition automatically. - Enforces strict selection and matching of digital human avatars and voices per task requirements. - Requires local ffmpeg tooling and proper credential management; no external LLM API keys needed for core flow. - Outputs MP4 with full workflow tracking and detailed JSON logs.
Metadata
Slug chanjing-one-click-video-creation
Version 1.0.10
License MIT-0
All-time Installs 1
Active Installs 1
Total Versions 11
Frequently Asked Questions

What is Chanjing One Click Video Creation?

用户输入选题或工作流,自动生成完整短视频成片(文案、分镜、数字人口播与 AI 画面混剪);调用 Chanjing Open API 与同仓库子技能脚本。 It is an AI Agent Skill for Claude Code / OpenClaw, with 245 downloads so far.

How do I install Chanjing One Click Video Creation?

Run "/install chanjing-one-click-video-creation" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Chanjing One Click Video Creation free?

Yes, Chanjing One Click Video Creation is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Chanjing One Click Video Creation support?

Chanjing One Click Video Creation is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Chanjing One Click Video Creation?

It is built and maintained by BinKes (@binkes); the current version is v1.0.10.

💬 Comments