diff --git a/server/mcp_server_mediakit/README.md b/server/mcp_server_mediakit/README.md index 21a918f4..dbe1022d 100644 --- a/server/mcp_server_mediakit/README.md +++ b/server/mcp_server_mediakit/README.md @@ -2,26 +2,16 @@ MediaKit MCP Server is a standard AI capability plugin for Volcano Engine AI MediaKit. It is built on the MCP (Model Context Protocol) protocol and exposes cloud media capabilities such as video editing, audio processing, subtitle processing, and video enhancement as tools that can be called by AI agents. With MediaKit MCP, developers can use natural language to drive intelligent media production workflows. -<<<<<<< HEAD -| Field | Value | -| ----------- | ----------------------------------------------------------------- | -| Version | v1.0.0 | -| Description | MediaKit MCP intelligent media assistant | -| Categories | Media cloud, audio/video editing, video enhancement | -| Tags | MCP, MediaKit, video editing, audio processing, video enhancement | -======= -| Field | Value | -| --- | --- | -| Version | v1.0.0 | -| Description | MediaKit MCP intelligent media assistant | -| Categories | Media cloud, audio/video editing, video enhancement | -| Tags | MCP, MediaKit, video editing, audio processing, video enhancement | - -> > > > > > > 62a9e293bc254026e04eaec63ca9d229aaa8e89e +| Field | Value | +| ----------- | ----------------------------------------------------------------------------------- | +| Version | v1.0.0 | +| Description | MediaKit MCP intelligent media assistant | +| Categories | Media cloud, audio/video editing, video enhancement, image processing | +| Tags | MCP, MediaKit, video editing, audio processing, video enhancement, image processing | ## Tool Overview -MediaKit MCP provides tools that cover the full workflow from asynchronous task query to deep media editing and video enhancement. All tools support dynamic loading by group or by tool name to optimize agent reasoning efficiency. +MediaKit MCP provides tools that cover the full workflow from asynchronous task query to deep media editing, video enhancement and understanding, audio processing, and image processing. All tools support dynamic loading by group or by tool name to optimize agent reasoning efficiency. @@ -42,8 +32,8 @@ MediaKit MCP provides tools that cover the full workflow from asynchronous task - - + + - - - - + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + @@ -133,7 +277,6 @@ Trae is an AI-native IDE with strong agent collaboration capabilities. By connec ## Prerequisites - Prepare a valid MediaKit API key. -- Confirm the MediaKit service endpoint. If not explicitly configured, the default is `https://amk.cn-beijing.volces.com`. - Install the [Trae client](https://www.trae.com.cn/). - For local mode or self-hosted cloud mode, make sure `uvx` is installed in your local environment. Run `uvx --version` to check. If it is not installed, follow the [uv installation guide](https://docs.astral.sh/uv/getting-started/installation/). @@ -143,18 +286,10 @@ Trae is an AI-native IDE with strong agent collaboration capabilities. By connec Choose one of the following modes based on your usage scenario: -<<<<<<< HEAD -| Mode | Best for | Access method | +| Mode | Best for | Access method | | --------------------------- | ------------------------------------------------------------------ | ---------------------------------------------------------------------------------------------- | -| **Local Mode (JSON Local)** | Personal debugging, quick trials, no self-hosted service required. | Use `uvx` to launch MediaKit MCP directly from the `mcp-server` repository subdirectory. | -| **Cloud Mode (JSON URL)** | Team sharing, long-term usage, centralized operations. | Deploy MediaKit MCP Server yourself, then connect using the deployed Streamable HTTP endpoint. | -======= -| Mode | Best for | Access method | -| --- | --- | --- | -| **Local Mode (JSON Local)** | Personal debugging, quick trials, no self-hosted service required. | Use `uvx` to launch MediaKit MCP directly from the `mcp-server` repository subdirectory. | -| **Cloud Mode (JSON URL)** | Team sharing, long-term usage, centralized operations. | Deploy MediaKit MCP Server yourself, then connect using the deployed Streamable HTTP endpoint. | - -> > > > > > > 62a9e293bc254026e04eaec63ca9d229aaa8e89e +| **Local Mode (JSON Local)** | Personal debugging, quick trials, no self-hosted service required. | Use `uvx` to launch MediaKit MCP directly from the `mcp-server` repository subdirectory. | +| **Cloud Mode (JSON URL)** | Team sharing, long-term usage, centralized operations. | Deploy MediaKit MCP Server yourself, then connect using the deployed Streamable HTTP endpoint. | ### Step 2: Add MCP Configuration @@ -178,8 +313,7 @@ Copy the following JSON and replace the fields as needed. Trae uses `uvx` to fet ], "env": { "MEDIAKIT_API_KEY": "your-api-key", - "MEDIAKIT_ENDPOINT": "https://amk.cn-beijing.volces.com", - "MCP_DOMAINS": "editing,video" + "MCP_DOMAINS": "editing,video,audio,image" } } } @@ -190,8 +324,7 @@ Copy the following JSON and replace the fields as needed. Trae uses `uvx` to fet - `mediakit_mcp`: The MCP service name. You can customize it. - `MEDIAKIT_API_KEY`: Replace with your MediaKit API key. -- `MEDIAKIT_ENDPOINT`: The MediaKit service endpoint. Keep `https://amk.cn-beijing.volces.com` if you use the default endpoint. -- `MCP_DOMAINS`: Load tools by group, for example `editing,video`. To load tools by exact tool name, use `MCP_TOOLS` instead. +- `MCP_DOMAINS`: Load tools by group, for example `editing,video,audio,image`. To load tools by exact tool name, use `MCP_TOOLS` instead. To load by tool name, use a configuration like this: @@ -207,7 +340,6 @@ To load by tool name, use a configuration like this: ], "env": { "MEDIAKIT_API_KEY": "your-api-key", - "MEDIAKIT_ENDPOINT": "https://amk.cn-beijing.volces.com", "MCP_TOOLS": "trim_video,query_task" } } @@ -223,7 +355,6 @@ A simple startup example is: ```bash export MEDIAKIT_API_KEY="your-api-key" -export MEDIAKIT_ENDPOINT="https://amk.cn-beijing.volces.com" export MCP_SERVER_HOST="0.0.0.0" export MCP_SERVER_PORT="8000" export STREAMABLE_HTTP_PATH="/mcp" @@ -240,8 +371,7 @@ After deployment, copy the following JSON and replace the fields as needed: "url": "https://your-domain/mcp", "headers": { "x-amk-api-key": "your-api-key", - "x-mediakit-endpoint": "https://amk.cn-beijing.volces.com", - "x-mcp-domains": "editing,video" + "x-mcp-domains": "editing,video,audio,image" } } } @@ -253,8 +383,7 @@ After deployment, copy the following JSON and replace the fields as needed: - `mediakit_mcp`: The MCP service name. You can customize it. - `url`: Replace with your self-hosted MediaKit MCP Streamable HTTP URL, such as `https://your-domain/mcp`. - `x-amk-api-key`: Replace with your MediaKit API key. -- `x-mediakit-endpoint`: The MediaKit service endpoint. Use `https://amk.cn-beijing.volces.com` if you use the default endpoint. -- `x-mcp-domains`: Load tools by group, for example `editing,video`. To load tools by exact tool name, use `x-mcp-tools` instead. +- `x-mcp-domains`: Load tools by group, for example `editing,video,audio,image`. To load tools by exact tool name, use `x-mcp-tools` instead. To load by tool name, use a configuration like this: @@ -265,7 +394,6 @@ To load by tool name, use a configuration like this: "url": "https://your-domain/mcp", "headers": { "x-amk-api-key": "your-api-key", - "x-mediakit-endpoint": "https://amk.cn-beijing.volces.com", "x-mcp-tools": "trim_video,query_task" } } @@ -312,16 +440,10 @@ The table below lists the core MediaKit MCP configuration fields for cloud mode - - - - - - - + @@ -335,85 +457,18 @@ The table below lists the core MediaKit MCP configuration fields for cloud mode For self-hosted cloud mode, you can also configure the following startup parameters: -<<<<<<< HEAD -| Environment variable | Default value | Description | +| Environment variable | Default value | Description | | ---------------------- | ------------- | --------------------------- | -| `MCP_SERVER_HOST` | `0.0.0.0` | MCP service bind address. | -| `MCP_SERVER_PORT` | `8000` | MCP service listening port. | -| `STREAMABLE_HTTP_PATH` | `/mcp` | Streamable HTTP path. | -======= -| Environment variable | Default value | Description | -| --- | --- | --- | -| `MCP_SERVER_HOST` | `0.0.0.0` | MCP service bind address. | -| `MCP_SERVER_PORT` | `8000` | MCP service listening port. | -| `STREAMABLE_HTTP_PATH` | `/mcp` | Streamable HTTP path. | - -> > > > > > > 62a9e293bc254026e04eaec63ca9d229aaa8e89e - -## Tool Details - -### query_task - -Query asynchronous task status. Supports one-shot query or polling through `poll_interval_seconds` and `max_poll_attempts`. - -### add_image_to_video - -Add an image overlay to a video. Commonly used for image watermarks. Supports image width, height, horizontal position, vertical position, start time, and end time. - -### add_subtitle_to_video - -Burn subtitle files or subtitle text into a video. Supports subtitle position, font size, font color, and font type. - -### adjust_video_speed - -Adjust video playback speed. Supports speed values from `0.1` to `4`. - -### concat_audio - -Concatenate multiple audio clips. Supports up to 100 audio URLs. - -### concat_video - -Concatenate multiple video clips. Supports up to 100 video URLs and optional transition effects. - -### extract_audio - -Extract audio from a video. Supports `mp3` or `m4a` output. - -### flip_video - -Flip video frames horizontally or vertically. - -### image_to_video - -Create an animated video from multiple images and optional transition effects. - -### mux_audio_video - -Combine a video and an audio file into one video. Supports preserving the original video audio and synchronizing duration by video or audio timeline. - -### trim_audio - -Trim an audio file by start and end time in seconds. - -### trim_video - -Trim a video by start and end time in seconds. - -### erase_video_subtitle_pro - -Remove subtitles or text from a video with high-quality restoration. Supports mainstream video formats such as `mp4`, `flv`, `ts`, `avi`, `mov`, `wmv`, and `mkv`. - -### enhance_video - -Enhance video quality for scenarios such as `common`, `ugc`, `short_series`, `aigc`, and `old_film`. Supports both standard and professional versions. +| `MCP_SERVER_HOST` | `0.0.0.0` | MCP service bind address. | +| `MCP_SERVER_PORT` | `8000` | MCP service listening port. | +| `STREAMABLE_HTTP_PATH` | `/mcp` | Streamable HTTP path. | ## License -MIT +This project is open-sourced under the **MIT License**. -This software calls MediaKit APIs at runtime. Use of these APIs is subject to the following terms and privacy policies: +At runtime, this software calls the MediaKit cloud APIs. Using these APIs is subject to the following agreements: -- [Video Cloud Service Special Terms](https://www.volcengine.com/docs/6448/79646?lang=zh) -- [Intelligent Processing Service Billing Rules](https://www.volcengine.com/docs/6448/104992?lang=zh) -- [Intelligent Processing Service Level Agreement](https://www.volcengine.com/docs/6448/79648?lang=zh) +- `https://www.volcengine.com/docs/6448/79646?lang=zh` +- `https://www.volcengine.com/docs/6448/104992?lang=zh` +- `https://www.volcengine.com/docs/6448/79648?lang=zh` diff --git a/server/mcp_server_mediakit/README_zh.md b/server/mcp_server_mediakit/README_zh.md index 62cd3466..2b1f9c1b 100644 --- a/server/mcp_server_mediakit/README_zh.md +++ b/server/mcp_server_mediakit/README_zh.md @@ -2,25 +2,16 @@ MediaKit MCP 是火山引擎 AI MediaKit 面向 AI 时代推出的标准能力插件。它基于 MCP（Model Context Protocol）协议，将云端专业的视频剪辑、音频处理、字幕处理、画质增强等原子能力封装为智能体可直观调用的工具。通过 MediaKit MCP，开发者可直接以自然语言驱动 AI 智能体完成复杂的云端媒体处理任务。 -<<<<<<< HEAD -| 字段 | 取值 | -| ---- | ------------------------------------------- | -| 版本 | v1.0.0 | -| 描述 | MediaKit MCP 智能媒体助手 | -| 分类 | 视频云、音视频编辑、画质增强 | -======= -| 字段 | 取值 | -| --- | --- | -| 版本 | v1.0.0 | -| 描述 | MediaKit MCP 智能媒体助手 | -| 分类 | 视频云、音视频编辑、画质增强 | - -> > > > > > > 62a9e293bc254026e04eaec63ca9d229aaa8e89e -> > > > > > > | 标签 | MCP、MediaKit、视频剪辑、音频处理、画质增强 | +| 字段 | 取值 | +| ---- | ----------------------------------------------------- | +| 版本 | v1.0.0 | +| 描述 | MediaKit MCP 智能媒体助手 | +| 分类 | 视频云、音视频编辑、画质增强、图像处理 | +| 标签 | MCP、MediaKit、视频剪辑、音频处理、画质增强、图像处理 | ## 工具概览 -MediaKit MCP 已开放的能力覆盖了从异步任务查询到深度媒体编辑、视频增强的全流程。所有工具均支持通过“分组（Group）”或“工具名”进行动态加载，以优化智能体的推理效率。 +MediaKit MCP 已开放的能力覆盖了从异步任务查询到深度媒体编辑、视频增强与理解、音频处理、图像处理的全流程。所有工具均支持通过“分组（Group）”或“工具名”进行动态加载，以优化智能体的推理效率。
Video editingeditingVideo editingediting add_image_to_video Add image to video: Overlay an image on a video, commonly used for image watermarks. For detailed input and output parameters, see add_image_to_video. @@ -110,17 +100,171 @@ MediaKit MCP provides tools that cover the full workflow from asynchronous task
Video enhancementvideoerase_video_subtitle_proErase video subtitles: Remove subtitles or text from a video with high-quality restoration. For detailed input and output parameters, see - erase_video_subtitle_pro. + adjust_audio_speedAdjust audio speed: Change audio playback speed for fast or slow effects. For detailed input and output parameters, see + adjust_audio_speed. +
adjust_video_volumeAdjust video volume: Change video volume, including muting. For detailed input and output parameters, see + adjust_video_volume. +
apply_video_filterApply video filter: Add a preset filter style to a video. For detailed input and output parameters, see + apply_video_filter. +
fade_audioFade audio: Apply fade-in and fade-out effects to an audio file. For detailed input and output parameters, see + fade_audio. +
fade_video_audioFade video audio: Apply fade-in and fade-out effects to a video's audio track. For detailed input and output parameters, see + fade_video_audio. +
mix_audioMix audio: Mix multiple audio files (background music, sound effects, vocals) into one audio file. For detailed input and output parameters, see + mix_audio. +
Video enhancement & understandingvideoanalyze_video_highlightsAnalyze video highlights: Detect emotional peaks and key actions and output highlight metadata (timestamps, scores, OCR text, scene descriptions). For detailed input and output parameters, see + analyze_video_highlights. +
analyze_video_storylineAnalyze video storyline: Parse film/TV content into a structured storyline of chronological clips and aggregated highlights. For detailed input and output parameters, see + analyze_video_storyline. +
asr_subtitlesSpeech to subtitles: Run speech recognition on a video or audio and output timestamped subtitle segments. For detailed input and output parameters, see + asr_subtitles.
enhance_video Enhance video: Improve video quality for AIGC, UGC, short drama, education, gaming, and old film restoration scenarios. For detailed input and output parameters, see - enhance_video. + enhance_video. +
enhance_video_generativeGenerative video restoration: Diffusion-model based restoration that reconstructs details and generates high-fidelity content. For detailed input and output parameters, see + enhance_video_generative. +
erase_video_subtitleErase video subtitles: Detect and erase existing hard subtitles while preserving the original background. For detailed input and output parameters, see + erase_video_subtitle. +
erase_video_subtitle_proErase video subtitles (pro): Remove subtitles or text from a video with high-quality restoration. For detailed input and output parameters, see + erase_video_subtitle_pro. +
generate_highlights_microdramaMicrodrama highlights: Extract highlights from microdramas and produce promotional cut videos. For detailed input and output parameters, see + generate_highlights_microdrama. +
generate_highlights_minigameMinigame highlights: Identify core gameplay and highlight events in minigame recordings and generate marketing video assets. For detailed input and output parameters, see + generate_highlights_minigame. +
matte_greenscreen_videoGreen screen matting: Matte videos with a green screen or solid-color background to produce a transparent-background video. For detailed input and output parameters, see + matte_greenscreen_video. +
matte_portrait_videoPortrait matting: Identify the human subject and remove the background to produce a transparent-background video. For detailed input and output parameters, see + matte_portrait_video. +
probe_video_metadataProbe video metadata: Probe a video URL and output standardized media metadata (container, video stream, audio stream). For detailed input and output parameters, see + probe_video_metadata. +
segment_scenesSegment scenes: Automatically split a video into scenes based on transitions and visual changes. For detailed input and output parameters, see + segment_scenes. +
video_ocrVideo OCR: Recognize subtitles/text in video frames and output timestamped subtitle segments. For detailed input and output parameters, see + video_ocr. +
Audio processingaudioseparate_voiceSeparate voice: Separate vocals and background sound into two independent tracks. For detailed input and output parameters, see + separate_voice. +
probe_audio_metadataProbe audio metadata: Retrieve detailed audio metadata (container and audio stream). For detailed input and output parameters, see + probe_audio_metadata. +
Image processingimageimage_ocrImage OCR: Recognize general printed text and return editable text, bounding boxes, and confidence. For detailed input and output parameters, see + image_ocr. +
erase_imageErase image: Detect and erase icons, text, or specified regions with intelligent background inpainting. For detailed input and output parameters, see + erase_image. +
remove_image_backgroundRemove image background: Keep the subject and remove the background to produce a transparent-background image. For detailed input and output parameters, see + remove_image_background. +
enhance_imageEnhance image: Improve image resolution, sharpness, and color based on content understanding. For detailed input and output parameters, see + enhance_image. +
evaluate_image_qualityEvaluate image quality: Score image quality and aesthetics (subjective and objective). For detailed input and output parameters, see + evaluate_image_quality.
your-api-key MediaKit API key used for authentication.
x-mediakit-endpointMEDIAKIT_ENDPOINThttps://amk.cn-beijing.volces.comMediaKit service endpoint. The default value is this endpoint.
x-mcp-domains MCP_DOMAINSediting,videoediting,video,audio,image Load tools by group. Separate multiple groups with commas.
@@ -41,8 +32,8 @@ MediaKit MCP 已开放的能力覆盖了从异步任务查询到深度媒体编 - - + + - - - - + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + @@ -132,7 +277,6 @@ Trae 是一款 AI 原生 IDE，提供了强大的智能体协作能力。通过 ## 前提条件 - 已准备可用的 MediaKit API Key。 -- 已确认 MediaKit 服务接入地址。若未显式配置，默认使用 `https://amk.cn-beijing.volces.com`。 - 已安装 [Trae 客户端](https://www.trae.com.cn/)。 - 使用本地模式或云端自部署模式时，需确保本地开发环境已安装 `uvx`。可通过 `uvx --version` 检查；若提示未安装，请参考 [uv 官方安装文档](https://docs.astral.sh/uv/getting-started/installation/)。 @@ -142,18 +286,10 @@ Trae 是一款 AI 原生 IDE，提供了强大的智能体协作能力。通过 根据您的使用场景，选择以下两种接入模式之一： -<<<<<<< HEAD -| 模式 | 适用场景 | 接入方式 | +| 模式 | 适用场景 | 接入方式 | | -------------------------- | ---------------------------------- | ------------------------------------------------------------------------- | -| **本地模式（JSON Local）** | 个人调试、快速试用、无需自建服务。 | 通过 `uvx` 直接从 `mcp-server` 仓库子目录拉起 MediaKit MCP。 | -| **云端模式（JSON URL）** | 团队共享、长期稳定使用、统一运维。 | 先自行部署 MediaKit MCP Server，再使用部署后的 Streamable HTTP 地址接入。 | -======= -| 模式 | 适用场景 | 接入方式 | -| --- | --- | --- | -| **本地模式（JSON Local）** | 个人调试、快速试用、无需自建服务。 | 通过 `uvx` 直接从 `mcp-server` 仓库子目录拉起 MediaKit MCP。 | -| **云端模式（JSON URL）** | 团队共享、长期稳定使用、统一运维。 | 先自行部署 MediaKit MCP Server，再使用部署后的 Streamable HTTP 地址接入。 | - -> > > > > > > 62a9e293bc254026e04eaec63ca9d229aaa8e89e +| **本地模式（JSON Local）** | 个人调试、快速试用、无需自建服务。 | 通过 `uvx` 直接从 `mcp-server` 仓库子目录拉起 MediaKit MCP。 | +| **云端模式（JSON URL）** | 团队共享、长期稳定使用、统一运维。 | 先自行部署 MediaKit MCP Server，再使用部署后的 Streamable HTTP 地址接入。 | ### 步骤 2：添加 MCP 配置 @@ -177,8 +313,7 @@ Trae 是一款 AI 原生 IDE，提供了强大的智能体协作能力。通过 ], "env": { "MEDIAKIT_API_KEY": "your-api-key", - "MEDIAKIT_ENDPOINT": "https://amk.cn-beijing.volces.com", - "MCP_DOMAINS": "editing,video" + "MCP_DOMAINS": "editing,video,audio,image" } } } @@ -189,8 +324,7 @@ Trae 是一款 AI 原生 IDE，提供了强大的智能体协作能力。通过 - `mediakit_mcp`：MCP 服务名称，您可以根据需要自定义。 - `MEDIAKIT_API_KEY`：请替换为您的 MediaKit API Key。 -- `MEDIAKIT_ENDPOINT`：MediaKit 服务地址。若使用默认接入地址，可保留 `https://amk.cn-beijing.volces.com`。 -- `MCP_DOMAINS`：按分组加载工具，例如 `editing,video`。如需按工具名精确加载，可改用 `MCP_TOOLS`。 +- `MCP_DOMAINS`：按分组加载工具，例如 `editing,video,audio,image`。如需按工具名精确加载，可改用 `MCP_TOOLS`。 如需按工具名加载，可参考以下写法： @@ -206,7 +340,6 @@ Trae 是一款 AI 原生 IDE，提供了强大的智能体协作能力。通过 ], "env": { "MEDIAKIT_API_KEY": "your-api-key", - "MEDIAKIT_ENDPOINT": "https://amk.cn-beijing.volces.com", "MCP_TOOLS": "trim_video,query_task" } } @@ -222,7 +355,6 @@ Trae 是一款 AI 原生 IDE，提供了强大的智能体协作能力。通过 ```bash export MEDIAKIT_API_KEY="your-api-key" -export MEDIAKIT_ENDPOINT="https://amk.cn-beijing.volces.com" export MCP_SERVER_HOST="0.0.0.0" export MCP_SERVER_PORT="8000" export STREAMABLE_HTTP_PATH="/mcp" @@ -239,8 +371,7 @@ uvx --from "git+https://github.com/volcengine/mcp-server.git#subdirectory=server "url": "https://your-domain/mcp", "headers": { "x-amk-api-key": "your-api-key", - "x-mediakit-endpoint": "https://amk.cn-beijing.volces.com", - "x-mcp-domains": "editing,video" + "x-mcp-domains": "editing,video,audio,image" } } } @@ -252,8 +383,7 @@ uvx --from "git+https://github.com/volcengine/mcp-server.git#subdirectory=server - `mediakit_mcp`：MCP 服务名称，您可以根据需要自定义。 - `url`：请替换为您自行部署后的 MediaKit MCP Streamable HTTP 地址，例如 `https://your-domain/mcp`。 - `x-amk-api-key`：请替换为您的 MediaKit API Key。 -- `x-mediakit-endpoint`：MediaKit 服务地址。若使用默认接入地址，可填写 `https://amk.cn-beijing.volces.com`。 -- `x-mcp-domains`：按分组加载工具，例如 `editing,video`。如需按工具名精确加载，可改用 `x-mcp-tools`。 +- `x-mcp-domains`：按分组加载工具，例如 `editing,video,audio,image`。如需按工具名精确加载，可改用 `x-mcp-tools`。 如需按工具名加载，可参考以下写法： @@ -264,7 +394,6 @@ uvx --from "git+https://github.com/volcengine/mcp-server.git#subdirectory=server "url": "https://your-domain/mcp", "headers": { "x-amk-api-key": "your-api-key", - "x-mediakit-endpoint": "https://amk.cn-beijing.volces.com", "x-mcp-tools": "trim_video,query_task" } } @@ -311,16 +440,10 @@ uvx --from "git+https://github.com/volcengine/mcp-server.git#subdirectory=server - - - - - - - + @@ -334,85 +457,18 @@ uvx --from "git+https://github.com/volcengine/mcp-server.git#subdirectory=server 云端自部署时，还可按需使用以下服务启动参数： -<<<<<<< HEAD -| 环境变量 | 默认值 | 说明 | +| 环境变量 | 默认值 | 说明 | | ---------------------- | --------- | ---------------------- | -| `MCP_SERVER_HOST` | `0.0.0.0` | MCP 服务监听地址。 | -| `MCP_SERVER_PORT` | `8000` | MCP 服务监听端口。 | -| `STREAMABLE_HTTP_PATH` | `/mcp` | Streamable HTTP 路径。 | -======= -| 环境变量 | 默认值 | 说明 | -| --- | --- | --- | -| `MCP_SERVER_HOST` | `0.0.0.0` | MCP 服务监听地址。 | -| `MCP_SERVER_PORT` | `8000` | MCP 服务监听端口。 | -| `STREAMABLE_HTTP_PATH` | `/mcp` | Streamable HTTP 路径。 | - -> > > > > > > 62a9e293bc254026e04eaec63ca9d229aaa8e89e - -## 工具详情 - -### query_task - -查询异步任务状态。支持单次查询，也支持通过 `poll_interval_seconds` 与 `max_poll_attempts` 控制轮询。 - -### add_image_to_video - -在视频中添加图片覆盖层，可用于图片水印。支持配置图片宽高、水平位置、垂直位置、开始时间和结束时间。 - -### add_subtitle_to_video - -将字幕文件或字幕文本压制到视频画面中。支持字幕位置、字体大小、字体颜色和字体类型配置。 - -### adjust_video_speed - -调整视频播放速度，实现快放或慢放效果。支持 `0.1` 到 `4` 倍速。 - -### concat_audio - -拼接多个音频片段，最多支持 100 个音频 URL。 - -### concat_video - -拼接多个视频片段，最多支持 100 个视频 URL，并可配置转场效果。 - -### extract_audio - -从视频中提取音频，支持输出 `mp3` 或 `m4a`。 - -### flip_video - -对视频画面进行水平或垂直镜像翻转。 - -### image_to_video - -将多张图片合成为动画视频，并支持配置转场效果。 - -### mux_audio_video - -将视频与音频合成为一个视频文件。支持保留原视频音频，并支持按视频或音频基准进行时长对齐。 - -### trim_audio - -按秒级起止时间裁剪音频文件。 - -### trim_video - -按秒级起止时间裁剪视频文件。 - -### erase_video_subtitle_pro - -针对视频中的字幕或文本进行高质量擦除，尽可能还原视频画面。支持 `mp4`、`flv`、`ts`、`avi`、`mov`、`wmv`、`mkv` 等主流视频格式。 - -### enhance_video - -针对 `common`、`ugc`、`short_series`、`aigc`、`old_film` 等场景进行画质增强，支持标准版和专业版工具版本。 +| `MCP_SERVER_HOST` | `0.0.0.0` | MCP 服务监听地址。 | +| `MCP_SERVER_PORT` | `8000` | MCP 服务监听端口。 | +| `STREAMABLE_HTTP_PATH` | `/mcp` | Streamable HTTP 路径。 | -## License +## 许可证 -MIT +本项目基于 **MIT 许可证** 开源。 -该软件运行时会调用 MediaKit 的 API，使用这些 API 需要遵守如下协议和隐私政策： +该软件运行时会调用 MediaKit 云端 API，使用这些 API 需要遵守如下协议： -- [视频云服务专用条款](https://www.volcengine.com/docs/6448/79646?lang=zh) -- [智能处理服务计费结算规则](https://www.volcengine.com/docs/6448/104992?lang=zh) -- [智能处理服务等级协议](https://www.volcengine.com/docs/6448/79648?lang=zh) +- `https://www.volcengine.com/docs/6448/79646?lang=zh` +- `https://www.volcengine.com/docs/6448/104992?lang=zh` +- `https://www.volcengine.com/docs/6448/79648?lang=zh`
视频剪辑editing视频剪辑editing add_image_to_video 视频加图片：在视频画面上叠加图片，常用于添加图片水印。详细输入和输出参数请见 add_image_to_video。 @@ -109,17 +100,171 @@ MediaKit MCP 已开放的能力覆盖了从异步任务查询到深度媒体编
视频增强videoerase_video_subtitle_pro视频字幕擦除：针对视频中的字幕或文本进行高质量无痕擦除。详细输入和输出参数请见 - erase_video_subtitle_pro。 + adjust_audio_speed音频播放调速：调整音频播放速度，实现快放或慢放效果。详细输入和输出参数请见 + adjust_audio_speed。 +
adjust_video_volume视频音量调整：调整视频音量大小，支持静音。详细输入和输出参数请见 + adjust_video_volume。 +
apply_video_filter视频滤镜：为视频添加指定的预设滤镜效果。详细输入和输出参数请见 + apply_video_filter。 +
fade_audio音频淡入淡出：对输入音频实现淡入淡出效果。详细输入和输出参数请见 + fade_audio。 +
fade_video_audio视频声轨淡入淡出：对输入视频的声轨实现淡入淡出效果。详细输入和输出参数请见 + fade_video_audio。 +
mix_audio音频混音：将多个音频文件（背景音乐、音效、人声等）混音生成一个新的音频文件。详细输入和输出参数请见 + mix_audio。 +
视频增强与理解videoanalyze_video_highlights视频高光分析：智能捕捉视频情绪波峰与关键动作，输出时间戳、高光打分、OCR 文本和画面描述等元数据。详细输入和输出参数请见 + analyze_video_highlights。 +
analyze_video_storyline视频剧情线分析：智能解析影视剧内容，生成由时序剧情片段与聚合高光组成的结构化剧情线。详细输入和输出参数请见 + analyze_video_storyline。 +
asr_subtitles语音识别字幕：对输入视频或音频进行语音识别，输出带时间戳的字幕片段。详细输入和输出参数请见 + asr_subtitles。
enhance_video 画质增强：面向 AIGC、UGC、短剧、教育、游戏、老片修复等场景提升视频画质。详细输入和输出参数请见 - enhance_video。 + enhance_video。 +
enhance_video_generative生成式视频修复：基于扩散大模型的生成式视频修复技术，补全细节、生成高保真视频内容。详细输入和输出参数请见 + enhance_video_generative。 +
erase_video_subtitle视频字幕擦除：智能检测并擦除视频画面中已有的硬字幕，保留原始背景。详细输入和输出参数请见 + erase_video_subtitle。 +
erase_video_subtitle_pro视频字幕擦除（Pro）：针对视频中的字幕或文本进行高质量无痕擦除。详细输入和输出参数请见 + erase_video_subtitle_pro。 +
generate_highlights_microdrama短剧高光智剪：深度理解短剧角色、剧情与故事线，自动提取高光片段并混剪成投流视频。详细输入和输出参数请见 + generate_highlights_microdrama。 +
generate_highlights_minigame小游戏高光智剪：识别小游戏录屏中的核心玩法与高光事件，快速生成买量视频素材。详细输入和输出参数请见 + generate_highlights_minigame。 +
matte_greenscreen_video绿幕抠像：对绿幕或纯色背景视频抠图，移除背景生成背景透明的视频。详细输入和输出参数请见 + matte_greenscreen_video。 +
matte_portrait_video人像抠像：自动识别人物主体并移除背景，生成背景透明的视频。详细输入和输出参数请见 + matte_portrait_video。 +
probe_video_metadata视频元信息探测：探测视频 URL，输出标准化媒资元信息（容器层、视频流层、音频流层）。详细输入和输出参数请见 + probe_video_metadata。 +
segment_scenes场景切分：依据视频转场与画面变化自动切分场景，输出切片时间轴和（可选）切片文件。详细输入和输出参数请见 + segment_scenes。 +
video_ocr视频 OCR：识别视频画面中的字幕/文字内容，输出带时间戳的字幕片段。详细输入和输出参数请见 + video_ocr。 +
音频处理audioseparate_voice人声分离：将音频中的人声与背景音精准分离，输出为两个独立音轨。详细输入和输出参数请见 + separate_voice。 +
probe_audio_metadata音频元信息探测：获取音频详细元信息，输出容器层信息与音频流元信息。详细输入和输出参数请见 + probe_audio_metadata。 +
图像处理imageimage_ocr图片 OCR：识别图片中的通用印刷体文字，返回可编辑文本、文字框坐标和置信度。详细输入和输出参数请见 + image_ocr。 +
erase_image图像擦除：自动检测并擦除图片中的图标、文字或指定区域，并对擦除区域进行背景智能填充。详细输入和输出参数请见 + erase_image。 +
remove_image_background图像背景移除：自动识别并保留图像主体，移除背景生成背景透明的图片，支持通用、人像、商品场景。详细输入和输出参数请见 + remove_image_background。 +
enhance_image图像画质增强：基于图像内容理解智能决策，提升图片分辨率、清晰度与色彩表现。详细输入和输出参数请见 + enhance_image。 +
evaluate_image_quality图像质量评估：对图片进行主客观画质与美学评分，支持标准版多维评分与专业版大模型评分。详细输入和输出参数请见 + evaluate_image_quality。
your-api-key MediaKit API Key，用于请求鉴权。
x-mediakit-endpointMEDIAKIT_ENDPOINThttps://amk.cn-beijing.volces.comMediaKit 服务地址；未配置时默认使用该地址。
x-mcp-domains MCP_DOMAINSediting,videoediting,video,audio,image 按工具分组加载，多个分组用英文逗号分隔。