WorkBuddy Automated Workflow Test: An 18-Step Review from Storyboard Script to MP4 Output

A no-holds-barred tool test recap: the input is 3 sets of natural-language prompts, the output is 4 3D storyboard images plus a vertical MP4 with voiceover and subtitles. This article reconstructs in 18 steps how WorkBuddy autonomously plans the workflow, self-heals missing environments, does frame-by-frame QC, and fixes the drawtext subtitle missing-glyph issue. At the end it also includes the step that didn't play smoothly.

I'm Xiaohu, a computer science grad from Zhejiang University, an "action breaker," and author of "AI Super Individual – The Complete Guide to One-Person Company." I focus on AI training, building super individuals, and helping companies improve efficiency.

Previously, to make a product video, an ordinary AI could only help you write storyboard copy. You still had to manually switch between separate software for image generation, voiceover, and editing.

WorkBuddy, from Tencent, can finally directly read and write local files, install its own environments, and call tools.

Hello, I'm Xiaohu.

I sent the requirement to it, and in 14 minutes it completed the entire pipeline from four 3D storyboard images, Xiaoxiao's female voiceover, to automatically fixing the subtitle garbling. Even though local playback stuttered at the end, this automated workflow is worth a look for small business owners.

Why is this worth reading now?

Many of my students always ask me: there are so many AIs on the market, why does making a video still feel exhausting?

According to Tencent's official documentation and the open-source use-case library maintained by EvoLinkAI, WorkBuddy is positioned as a digital employee that can autonomously plan complex workflows and actually execute tasks.

It's not just a chat companion; it supports multi-model collaboration and the MCP ecosystem, and can read and write local files, execute code, and call tools on its own. That lets one person stand in for a whole team of writers, editors, and designers – a "one-person company" (OPC).

For small merchants who don't have the budget to hire a professional video outsourcing team, the difference is very intuitive:

Dimension	Ordinary AI dialogue tool	WorkBuddy automated workflow
Execution ability	Can only give copy suggestions, doesn't touch the local environment	Automatically installs missing environments (e.g., edge-tts), calls tools
File read/write	Can't directly write to the system, requires manual copy-paste	Automatically creates and reads local files, generates MP4 and storyboard images
Task span	Single-step dialogue, stops after each segment	Plans multi-step tasks, from writing the script to video rendering in one go
Deliverable	Pure text replies, editing still on you	Delivers an accept-ready finished video, with multi-frame QC and self-healing

How to do it

Step 1: Create a storage folder on the desktop

On the desktop, create a new folder named 短剧实测 (Short Drama Test) to store all subsequent storyboard images and video files.

Step 2: Open the client and start a new session

Open the WorkBuddy desktop client, click the "New Task" button in the top-left to start a fresh conversation, and collapse the history sidebar on the left to keep the workspace clean.

Step 3: Send the storyboard script generation command

Enter the storyboard requirement in the bottom input box. During this test, the first sentence accidentally started with "11快到了" (11 is coming), which didn't affect understanding.

Prompt:

11快到了，我想做一条30秒的智能空气炸锅带货动画短剧，手机竖屏看的。帮我写个3到4个镜头的分镜脚本，要有剧情冲突、角色动作、空气炸锅特写和台词，不要凑篇幅的废话。

Translation: "11 is almost here, I want to make a 30-second animated short drama to promote a smart air fryer, for vertical mobile screens. Write a 3–4 shot storyboard script with plot conflict, character actions, close-up of the air fryer, and dialogue. Don't pad with useless words."

Step 4: Confirm storyboard script and consumption

After sending, wait briefly and close extra embedded pages so the chat area centers. WorkBuddy replied with "11.11 Eve, Kitchen Battle", a 30-second vertical, 4 shots. This step consumed 1.56 credits, and it saved the full script as a local Markdown file.

I've coached many students; they usually hate long pre-script procedures. Having it break down the 4-shot rhythm directly from a command is genuinely time-saving.

Step 5: Query available free production options

Paste the second prompt in the input box. The screenshot shows the input box with messy repeated edits, but the real core requirements are these lines:

Prompt:

看一下你现在有哪些能免费用的在线AI画图、动态运镜和配音合成工具或技能？列出来给我看看，没有的话帮我装个好用的，千万别用代码画简陋的几何图形凑数。

Translation: "Check what free online AI image, dynamic camera movement, and voiceover tools or skills you have now. List them for me; if not available, install a good one. Don't use code to draw crude geometric shapes as filler."

Step 6: Wait for environment search

After clicking send, WorkBuddy automatically searches the local and network for available free audio-video tool environments; the interface shows it's searching local voices and files.

Step 7: Enter storyboard generation stage

Check WorkBuddy's tool inventory results; it starts automatically calling the built-in drawing capability and enters 3D animation movie style storyboard generation.

Step 8: Wait for 3D storyboard rendering

WorkBuddy calls AI image generation over the network to produce Pixar-quality 3D cartoon sequential storyboard images. The interface estimates 1–5 minutes for drawing.

Step 9: Check character and scene details

In the right-hand preview area, view the auto-generated high-resolution 3D cartoon storyboard frames. The first frame shows a woman coming home from overtime, one hand holding a bag, the other on the phone, with a oil pan smoking on the gas stove – very lively, nothing like the old stiff flat characters.

Step 10: Storyboard archiving and QC acceptance

WorkBuddy indicates "All 4 frames hold up; the Chinese text on screens in frames 2 and 4 – '智能空气炸锅 / 180℃ / 25分钟' – is all rendered correctly." It auto-completes fields and archives.

Step 11: Verify the tool list and asset table

Browse the capability inventory table and storyboard mapping table in the conversation area to confirm all four storyboard images are ready.

Step 12: Send the composite video command

Paste the fourth prompt in the bottom input box and click send:

Prompt:

现在给这几张分镜图配上自然的中文配音和同步字幕，再加上平滑的推拉运镜特效，合成一条30秒的竖屏MP4短片，也存进桌面「短剧实测」文件夹，做完把完整文件路径和耗时告诉我。

Translation: "Now add natural Chinese voiceover and synced subtitles to these storyboard images, plus smooth push-pull camera effects, and composite a 30-second vertical MP4 video. Save it to the desktop '短剧实测' folder, and when done tell me the full file path and time taken."

Step 13: Wait for the pipeline to build

After sending, WorkBuddy goes into deep thinking, auto-calls speech synthesis and video editing components to build the rendering pipeline.

Step 14: Auto-install environment and voiceover

WorkBuddy detects that edge-tts is missing from the system, installs and configures it on its own. Then it completes the voiceover generation (4 lines totaling 15.8 seconds, Xiaoxiao female voice at +8% speed, natural timbre), and calls scripts for push-pull camera movement and subtitle burning.

Step 15: QC error detection and auto-fix

After the initial composite, WorkBuddy runs auto-QC and finds a flaw: "At subtitle line breaks, '口' boxes appear – on Windows, text files default to \r\n line endings, and drawtext renders the carriage return as a missing-glyph box." It then automatically investigates and modifies the script, re-rendering.

Step 16: Re-composite and archive request

After the video is re-composited, the system prepares to unify names for the 4 storyboard images and archive them to the desktop. At this point, the right embedded player says the video encoding format isn't supported for browser preview; you need to download it and use a local player. The conversation area is about to submit the final archive confirmation.

Step 17: Accept final delivery list and time

WorkBuddy then finishes up and reports: the final MP4 is vertical, 13.7 MB, plus 4 storyboard images from 2.0 MB to 2.4 MB, all archived to the desktop 短剧实测 folder.

The composite script ran for 12.7 seconds; the whole flow – "receive order → self-heal and install edge-tts → voiceover 10 seconds → composite → QC 3 frames frame-by-frame → fix 2 subtitle bugs → archive" – took about 14 minutes.

Step 18: Play the final video with a local player

Open the MP4 in the desktop 短剧实测 folder in fullscreen with a local player, press spacebar repeatedly to check the picture, voiceover, and subtitles. But the picture gets stuck at decoding/rendering; after multiple tries, it still doesn't play smoothly.

How good is the finished product?

Looking at the first finished frame, the overall image quality is solid:

The unexpected plus: high fidelity of Chinese characters in the images. All four 3D cartoon storyboard frames hold up, especially the "智能空气炸锅 / 180℃ / 25分钟" text on the air fryer screen in frames 2 and 4 – those Chinese characters actually render correctly, no common generating gibberish.

On the compositing side: as someone who's written code for over a decade, I'm impressed that it could autonomously figure out that Windows defaults to \r\n in text files, which caused subtitle line breaks to render as '口' missing-glyph boxes, then fix the script and rerun to self-heal. Plus the three-frame QC check aligns exactly with the "acceptance upfront" approach I always emphasize in deliveries.

But this test also had an awkward shortfall: the final video was produced, but the built-in preview warned from the start that the encoding format wouldn't play in a browser. When I opened it with the system's local player and hit space, the player kept looping in place because decoding/rendering couldn't keep up; after a lot of effort, it never played through smoothly.

If you want to keep improving after the basic video works, you can give it more tasks: ask it to increase speech speed, add product selling points in large type at the top, or have it directly generate Xiaohongshu and Douyin promotional copy based on the storyboard script.

For micro-merchants or side-hustle beginners who want to ride the Double 11 buzz for product animation but are on a tight budget and can't afford outsourcing, copying the prompts into WorkBuddy and letting it auto-configure free tools to run the short video pipeline is a completely actionable next step.

That's the full test record. All 3 prompt sets can be reused directly.