One Sentence in WorkBuddy: AI Assembles a 106-Second Long March Map Documentary via MCP

MCP Practice: In WorkBuddy, a single sentence lets AI generate a Red Army Long March map animation (Weijie AI Map Animation MCP integration tutorial)

Bottom line first: After connecting the "Weijie AI Map Animation" MCP in WorkBuddy, with only one sentence — "Call the map-video MCP tool, refer to platform cases, implement the Red Army Long March science popularization introduction, the effect should be good" — the AI read the API documentation, wrote scripts, verified place names, added music, and reviewed the film, all on real satellite maps. It produced a 106-second, eight-chapter Red Army Long March educational animation. From Ruijin to the Huining rendezvous, the 25,000-li route grows along the way.

The whole process required no CapCut, no AE, no having AI write thousands of lines of code from scratch, and no manually drawing a path. It took just 20–30 minutes, and WorkBuddy used under 40 credits. Below are clips from the final film and a complete 12-step tutorial from zero.

📺 Video version: The full operation process and the complete 106-second final film are in the Bilibili video: https://www.bilibili.com/video/BV1e64n6SEe4

How hard was making a Long March route video before?

For historical educational videos about the Long March, Zheng He's voyages, the Silk Road, Xuanzang's westward journey — an animated map is almost essential. Audiences need to see where the team started, where battles happened, which mountains were crossed, and where they finally arrived. But to actually make one yourself, you typically hit one of these three walls.

Route 1: CapCut — easy but no map

CapCut handles editing, subtitles, music, and templates smoothly, and you can produce a video in minutes. But it has no real maps:

• You can only paste a static map screenshot; zooming in blurs it.
• To make a route "walk," you have to trace it segment by segment on the screenshot, setting keyframes frame by frame. A full Long March route takes an entire day.
• Where is Ruijin? Zunyi? Luding Bridge? You have to locate them yourself on the map, and no one warns you if you trace it wrong.
• To change a map style or fix a place name, you start over from the screenshot.

Route 2: AE + GEOlayers 3 — possible but AI can't help

The professional solution for map animation is After Effects plus the GEOlayers 3 plugin: it loads online maps, does route animation, and looks good. The problems:

• High barrier: First learn AE, then the plugin. Layers, masks, expressions, and keyframes are all unavoidable.
• Fully manual: Finding coordinates, drawing routes, positioning cameras, adding labels, and setting pacing — every step is clicked by hand.
• AI is useless: No matter how capable AI is, it cannot operate AE and the plugin for you. You can get AI to write a Long March script, but you can't make AI turn that script into a map animation.
• Slow output, painful edits: With big projects, preview and rendering require waiting, and changing one node often pulls a lot of other things.

Route 3: Let AI tools directly write code to make the animation — expensive, time-consuming, and the result is mediocre

Since AI can write code, why not let tools like Cursor or Claude Code directly write a map animation? Many have tried, and the conclusion is similar:

• Time-consuming: AI must build the project from scratch, pick a map library, write route and camera animations, then run it, find it wrong, fix it, run again — a dozen rounds is common, and an afternoon disappears.
• Troublesome: You have to watch over dependency installation, project runs, and error debugging. Once the animation is written, you still have to record the screen or hook up ffmpeg to get a video, and subtitles and music are separate.
• Costly: Generating thousands of lines of code from zero plus round after round of modification consumes far more model usage than you'd expect.
• Mediocre result: The basemap, route, camera, subtitles, and music are all hastily assembled, often with stutters, glitches, and erratic pacing. Switching topics means starting over, and results vary every time.

The problem is not that AI is dumb; it's that you're making AI reinvent the wheel from scratch. Map engines, timelines, cameras, rendering — these underlying capabilities should never be rewritten for every single video.

The fourth path: Let AI produce directly on a real map

Weijie AI Map Animation is an AI video editor that uses real maps as its canvas: timelines, subtitles, and music work like a normal editor, but the canvas is a real map. It provides a standard MCP interface. After connecting AI assistants like WorkBuddy, Cursor, or Claude Code, they can directly create projects, write scripts, add routes, add music, review the film, and export the video.

The biggest difference from "AI writing code from scratch" is that maps, cameras, timelines, music, and rendering are all pre-built in the platform. AI only acts as the "scriptwriter and director." For this Long March film, the script AI wrote was only 7.7 KB, and the finished video came in 20–30 minutes with fewer than 40 WorkBuddy credits.

	CapCut etc.	AE + GEOlayers 3	AI tools write code directly	WorkBuddy + Weijie AI Map Animation MCP
Real maps	No, only screenshots	Yes	AI must integrate map libraries	Satellite imagery, terrain, administrative divisions ready to use
Route animation	Hand-trace screenshots, set keyframes	Possible, fully manual	Possible, repeated debugging	AI generates routes, camera moves, and labels
Place name accuracy	You verify everything yourself	Accurate, but you find the coordinates	Depends on AI's luck	Anchored to real coordinates, AI dry-runs to verify
Turning into video	Direct export	Slow rendering	Record screen or wire up ffmpeg	One-click export, up to 4K
Time and cost	A day of tracing	Weeks for learning and production	Several hours of back-and-forth, heavy model usage	20–30 minutes, under 40 credits (WorkBuddy gives 100 free credits daily)
Editing afterward	Redo from screenshot	One change often affects many things	Change code, often introduces new bugs	Continue the conversation, or edit frame-by-frame on the timeline
Onboarding difficulty	Low	High	Medium, must know how to run projects	Copy configuration + say one sentence

(The table reflects the typical situation for each approach.)

Why "WorkBuddy + Weijie AI Map Animation MCP"

In this practical test, this combination offered several tangible benefits:

1. The map is real, not "drawn." The basemap is real satellite imagery. Ruijin, Zunyi, Luding Bridge, and Wuqi Town all sit on real coordinates, and the route follows real terrain. The last thing an educational video needs is a wrong map — this is the bottom line.
2. One sentence produces a "project," not a dead video. AI generates a complete map animation project: every segment, every label, and every subtitle sits on the timeline, and the script file stays in your local folder. To change something, just continue the conversation, or open the editor and drag things yourself.
3. AI "learns techniques by looking at cases." The platform has 70+ public cases. Add "refer to platform cases" to your prompt, and AI first studies how others do opening credits, structure chapters, and style, then gets to work.
4. AI reviews its own work. Before final export, AI runs a "dry run" on the server side, verifying place names and checking camera and subtitle timing. This time the review result was 0 errors, 0 warnings.
5. Audio and visuals in one stop. Aged color grading, vignetting, film grain, a red flowing route, documentary-style subtitles, march music, plus bugle and war drum sound effects — all done within one project, no need for other software.
6. Flexible export. Server-side rendering, browser rendering with your local GPU, or desktop rendering all work. Switch between 1080p, 4K, Douyin vertical, and Bilibili 60fps with one click.
7. Saves time and money. The underlying capabilities are all ready in the platform. AI doesn't write thousands of lines of code; it only writes a few-KB script, and the film appears in 20–30 minutes. This entire video used fewer than 40 WorkBuddy credits, far less than making AI write code from scratch.
8. Works with any AI tool. MCP is an open standard. WorkBuddy, Cursor, Claude Code — all can connect. This test even used a light and fast model like DeepSeek-V4.1-Flash, and the results were just as good.

Hands-on: 12 steps to make a Red Army Long March educational animation

Overall, it's four things: copy configuration → connect to WorkBuddy → say one sentence → export the final film. Step by step below.

Step 1: Open Weijie AI Map Animation

In a browser, open vjmap.com/agent/ext/map-video/#/projects and log in with your Weijie Map account (shared with the official site; register if you don't have one). You'll see a project list. On the left, "Platform Cases" has over 70 public examples, covering timelines, map editing, cameras, and data visualization. AI will reference these cases when making your film.

Step 2: Copy the MCP configuration

Click the "MCP" button in the upper right. The "Connect Cursor · Project Library" panel appears. Click "Copy mcp.json configuration".

This "project library address" is not bound to any single project: once AI connects, it can list all your projects, pick one to edit, or create a new one.

⚠️ Note: This address equals your account permission (it can read, modify, and delete all your projects). Only put it into your own AI tools; don't share it. If it leaks, click "Reset address" in the panel and the old address immediately becomes invalid. To show your work to others, use "Share."

Step 3: Switch WorkBuddy to development mode and choose a local folder

Open WorkBuddy, create a new task, switch the mode at the top to "Development", then in "Select workspace" below the input box choose a local folder (here we created a video folder). AI-written scripts and assets will go there, making later edits easy.

Step 4: Click "+" → Connector → Manage connectors

Click the "+" in the lower left of the input box, choose "Connector" from the menu, then click "Manage connectors" at the bottom of the list.

Step 5: Click "Custom connector"

At the upper right of the connector page, click "Custom connector".

Step 6: Add MCP and paste the configuration you copied

In "MCP Service Management", click "Add MCP", paste the mcp.json configuration from Step 2, and click "Save". The configuration contains only one map-video service and its address; nothing needs to change.

At this point, WorkBuddy "learns" how to make map animations.

Step 7: Ask with one sentence

Back in the new task, type your requirement in the input box. A recommended template:

Call the map-video mcp tool, refer to platform cases, implement xxx feature

This time we entered:

Call the map-video mcp tool, refer to platform cases, implement the "Red Army Long March science popularization" introduction, the effect should be good

Two other settings in the screenshot are worth explaining:

• Model: This time we chose DeepSeek-V4.1-Flash; other models work too.
• Permission: We selected "Allow full access." AI can run commands and write files without asking for permission at every step, suitable for letting it finish a whole task. If you prefer to review every step, keep the default permission.

Step 8: WorkBuddy gets to work

After you send it, you can go make tea. WorkBuddy will first read the map animation's interface description to understand what the platform can do and how to write the script. Then it creates a film directory, writes the script, runs a server-side dry run, verifies place names, and reviews the film — when it finds a problem, it fixes it and runs again.

The whole process is a complete "scriptwriting + editing + review" loop. You don't need to intervene.

Step 9: Refresh the map animation page, the project is already built

While it works, go back to the browser and refresh the Weijie AI Map Animation project list. A new project "Red Army Long March" will already be there, with its duration slowly changing and the cover generating — AI is "editing" inside.

Step 10: WorkBuddy finishes

After 20–30 minutes and fewer than 40 credits, WorkBuddy reports completion and also writes down its "director's notes":

• Visual: aged color grading on the satellite basemap, vignette, film grain, red flowing route, giving a documentary feel.
• Subtitles: documentary style.
• Music: "March (military march)" from the library, automatically lowered when narration appears.
• Review: 0 errors, 0 warnings, with only one hint about "ending narration crossing a camera cut point" that doesn't actually affect viewing.
• Files: The script film.mv.ts remains in your local folder, and the project can be edited at any time.

It also shows several keyframe previews and suggests the next step: open the project page, in the AI assistant's "MCP" panel at the top enable "Allow external AI to operate this page," reply "Done," and it will begin exporting 1080p.

Step 11: Export the final film — if the server is slow, render locally in the browser

There are two export methods:

• Server export: Click "Export" in the upper right of the editor and let the server render. You can walk away. If the server is busy with many tasks, it may be slow.
• Browser export on your machine: Follow WorkBuddy's prompt, enable "Allow operation on this page" in the editor's "MCP" panel, and WorkBuddy can initiate the export directly in this page, rendering frame by frame with your computer's GPU without occupying the server, and the final video is automatically uploaded back to the platform. Generally much faster than server export.

Tip: During browser export, keep this page in the foreground. If you switch away or minimize, the browser may slow down or pause rendering.

Step 12: Install the desktop client for faster local export

If you export often, consider installing the Weijie Map Animation desktop client (Windows / macOS). The desktop client uses the same editor, account, and projects as the web version. The difference is in output: it directly calls your local GPU to render, saving the final video locally.

In the export dialog, under "Output" select "This computer" (here it's an RTX 3050 laptop GPU). A 106-second, 3189-frame film is expected to render in 5–15 minutes. The specifications are comprehensive:

• Common: 1080p, 720p quick draft.
• Platforms: Douyin / WeChat Channels vertical (1080×1920), Bilibili 1080p60.
• High-end: 4K horizontal, 4K vertical.
• Post-production: Transparent WebM, ProRes, PNG sequence — ready for further work in AE or CapCut.
• Attachments: Cover frame, subtitle SRT / VTT exported together.

Before exporting, the dialog also does a health check. This time it warned, for example, "105.8–106.3 seconds have no camera." Clicking "add a hold camera" fixed it.

Desktop client download: vjmap.com/guide/map-video/desktop.html, available from both the official site and GitHub.

Result: AI's own eight-chapter structure

Open the project, and you can see the complete timeline AI built: camera keyframes, point markers for Ruijin and Wuqi Town, year and subtitle tracks, route, music, and sound effects each have their own track. Every segment can be manually fine-tuned afterward, just like a conventional editor.

The whole film is 106 seconds, divided into eight chapters along the timeline:

Chapter	Time	Visual
① Departure from Ruijin	1934.10	Opening title, camera pushes into Ruijin: 86,000 people set out on the long march
② Battle of Xiang River	1934.11	Route grows westward: the Red Army drops from 86,000 to over 30,000
③ Zunyi Conference	1935.01	A turning point of life and death
④ Four Crossings of the Chishui River	1935.01—03	Attack arrows weave back and forth, feint east and strike west
⑤ Forcing the Dadu River, seizing Luding Bridge	1935.05	22 warriors grip iron chains and take Luding Bridge
⑥ Climbing snowy mountains, crossing grasslands	1935.06—08	Crossing Jiajin Mountain over 4,000 meters, traversing vast marshy grassland
⑦ Arrival at Wuqi Town	1935.10	Central Red Army reaches northern Shaanxi
⑧ Huining Rendezvous	1936.10	The three main forces meet; camera pulls back, the entire 25,000-li route concludes

Here are a few clips to feel it. Four Crossings of the Chishui River: the red attack arrows weave between Sichuan, Guizhou, and Yunnan, making "feint east, strike west" clear at a glance:

Forcing the Dadu River, seizing Luding Bridge: the camera follows the route into the Hengduan Mountains, with real terrain of snowy peaks and gorges:

Each chapter has large year text, battle labels, and documentary subtitles. The route grows with a red flowing glow, paired with a march and bugle/drum sounds. It's suitable for classroom playback, party history study, and WeChat article illustrations.

Want a different topic? Just change one sentence

The prompt template is the same: call the map-video mcp tool, refer to platform cases, implement xxx feature. Some tips:

• Add "refer to platform cases": AI looks at public cases first, and the opening, chapters, and styling are noticeably better.
• Add "the effect should be good": AI runs the dry run and review more seriously, fixing problems itself.
• Specify duration and aspect ratio: e.g., "60 seconds, vertical, for Douyin" — AI arranges shots accordingly.
• If not satisfied, just keep talking: "Change the route to gold," "Hold on Four Crossings for 3 more seconds," "Add a narration line to the opening" — AI modifies the existing project without touching the rest.

Ready-to-try topics:

• "Call the map-video mcp tool, refer to platform cases, make a 60-second animation of Zheng He's seven voyages."
• "Call the map-video mcp tool, refer to platform cases, make a route animation of the Silk Road from Chang'an to Rome with subtitles."
• "Call the map-video mcp tool, refer to platform cases, make a vertical route animation of Xuanzang's westward journey for Douyin."
• "Call the map-video mcp tool, refer to platform cases, turn a self-driving route from Beijing to Xi'an into a 30-second travel vlog opening."

FAQ

Can it only be used with WorkBuddy? No. MCP is an open standard. Any AI tool supporting MCP, such as Cursor or Claude Code, can connect. In Cursor, you can also click "one-click add to Cursor" in the panel.

Can I still edit after AI finishes? Yes. The generated product is a normal map animation project. Open the editor and drag on the timeline, change text, recolor, add cameras. Or continue the conversation in WorkBuddy and let AI modify it.

Will the map be wrong? The basemap is real satellite imagery, and place names are anchored to real coordinates. AI runs a server-side dry run before export to verify place names and review the film. Still, for historical educational content, it's recommended you cross-check against historical sources before publishing.

Can it export without a discrete GPU? Yes. Server export doesn't need a local GPU. If you have a GPU, using the browser or desktop client for local export is faster.

Can it be privately deployed? Yes. Both the Weijie Map agent platform and AI Map Animation support intranet deployment. Contact us for a solution.

Closing words

In the past, making a map educational video meant either tracing lines frame-by-frame over screenshots in CapCut, spending weeks on AE and GEOlayers, or watching AI write code for hours and getting a disappointing result. Now, WorkBuddy handles thinking and doing, while Weijie AI Map Animation provides real maps and a professional timeline. You only need to say clearly what story you want to tell. In 20–30 minutes and under 40 credits, you have your finished film.

The 25,000-li Long March can be re-walked on a map with one sentence. Open vjmap.com/agent/ext/map-video, copy the MCP configuration, and try your first sentence.

---

• Start creating: https://vjmap.com/agent/ext/map-video/#/projects
• Product introduction: https://vjmap.com/products/map-video
• Documentation: https://vjmap.com/guide/map-video/
• Desktop client download: https://vjmap.com/guide/map-video/desktop.html
• Video version: https://www.bilibili.com/video/BV1e64n6SEe4
• Related reading: AI Map Animation – an AI video editor with real maps as the canvas