# metaontic-video **Repository Path**: hackloop/metaontic-video ## Basic Information - **Project Name**: metaontic-video - **Description**: No description available - **Primary Language**: Unknown - **License**: MIT - **Default Branch**: master - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2026-10-29 - **Last Updated**: 2026-11-03 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README

# MetaOnticVideo: Seedance AI Video Creation for DeepSeek Harness **Create AI videos longer than 15 seconds using Seedance 2.0 multi-segment video continuation.** Start with a natural-language idea, review the script and storyboard, generate and preview each segment, then merge the results locally with FFmpeg. [中文](./README.md) · [Installation and user guide (Chinese)](./MetaOnticVideo使用说明书.md) · [Latest release](https://github.com/wxsimon2022/metaontic-video/releases/latest) · [Changelog](./CHANGELOG.md) · [Gitee mirror](https://gitee.com/hackloop/metaontic-video) ## Why MetaOnticVideo? Seedance 2.0 generates up to 15 seconds per request. MetaOnticVideo plans a longer video as an initial generation followed by video continuation: a 30-second project can use two 15-second segments, and a 60-second project can use four. Preview and approve each segment before continuing; the plugin keeps the source segments and merges a complete local video. The project planner supports up to 120 seconds and 10 segments. Longer videos require multiple paid model calls. Transitions express the intended continuity; inspect each generated segment to assess the actual result. ## Features - **Text-to-video and image-to-video:** prompts, reference images, first frames or first/last frames, depending on the model and provider. - **Script and storyboard review:** review the full script, a local storyboard illustration and cost estimates before confirming production. - **Video continuation:** describe transitions and extend the previous video while retaining the approved subject, action and audio plan. - **Segment previews and resume:** review each completed segment and continue using the saved task state; reuse completed segments after an interruption. - **Local FFmpeg merging:** deliver the merged video first, followed by the original numbered segments. - **480P and 720P:** confirm resolution during script review; subsequent segments retain that choice. - **Provider options:** MetaOntic AI, a service-issued key, or direct Volcano Engine access. The current model selector offers Seedance 2.0 Mini and Fast. - **Optional character design:** for a text-only project, choose whether to generate a stylized character image with Seedream. Image calls are billed separately; review the result before using it for the initial video and all continuations. Declining keeps the text-only workflow. - **Reference image preparation:** supported relay connections can use Seedream to convert real-person references into a non-photorealistic style, with separate consent, billing and result review. ## Quick start 1. Open [the latest GitHub release](https://github.com/wxsimon2022/metaontic-video/releases/latest) or [Gitee Releases](https://gitee.com/hackloop/metaontic-video/releases). Copy the download link for the `metaontic-video-.tgz` asset and paste it into **DeepSeek Harness → Plugins → Add plugin**. Use the packaged asset rather than a Git repository or Source code archive. Alternatively, download and extract the asset, then install its local `package` directory. On Windows, `ERR_PNPM_GIT_RESOLVE_FAILED` indicates a Git source failed to resolve; switch to the packaged asset without changing global Git settings or disabling supply-chain checks. 2. Open the plugin settings, select **MetaOntic AI**, register at [ai.bileego.com](https://ai.bileego.com/), obtain an API key and save it in the plugin. A shared key must support both selected video and image models; separate keys are also supported. 3. Start a **new conversation** and select **视频创作助手 (Video Creation Assistant)**. 4. Describe your video, including duration, aspect ratio and resolution. Review the script and storyboard, then separately confirm production and costs. Example: ```text Create a 30-second video introducing a melon farm, 16:9, 480P. Start with the farmer introducing himself, then show him gently tapping a ripe melon. Use Mandarin dialogue and ambient sound, without subtitles. Show the complete script, storyboard and cost estimate before generating anything. ``` ## Requirements and workflow DeepSeek Harness provides the plugin runtime; Node.js 20+ is required. FFmpeg is needed for local multi-segment merging, and ffprobe is used to inspect media. See the [FFmpeg setup instructions](./README.md#ffmpeg-与设置页). The storyboard is a locally drawn illustration, not a paid image generation or a preview of the final model output. Reference-image conversion, retries and additional media calls are billed separately. Model and provider capabilities determine which inputs are accepted. If generation stops or a status query fails, check the original task ID or resume with the saved `resumeDir`. A missing response does not prove that the provider did not create a task; the plugin blocks automatic resubmission until the original request is checked. ## Documentation and support - [Full Chinese documentation](./README.md) - [Installation, key configuration and usage guide](./MetaOnticVideo使用说明书.md) - [User guide PDF](./MetaOnticVideo使用说明书.pdf) - [Local development and hot reload](./README.md#本地开发热加载) - [Report an issue](https://github.com/wxsimon2022/metaontic-video/issues) - [Download release packages](https://github.com/wxsimon2022/metaontic-video/releases) ## License [MIT](./LICENSE) Release packages bundle the required host schema library and do not require external runtime npm dependencies. If a previous local-development setup still points to a deleted source directory, correct that local loading path before installing a release package. Windows users can also download the `MetaOnticVideo-Windows-.zip` release asset, extract it and install the included `.tgz` file (the `package` directory is for documentation and recovery). The ZIP includes an illustrated Chinese installation tutorial and the full PDF guide. Do not paste the ZIP or release page URL into the installer. Before an online update, the plugin checks the recorded installation source. Directory links are blocked from in-place replacement; unavailable source information or metadata errors also stop installation. Managed tarball and repository sources remain supported. An already missing directory must be restored and its loading path checked separately.