New Media Content Factory · content on autopilot
A large topic library,
agents across nine angles,
content every day.
Snap a product photo, or pull topics from the library in order, at random or by trend. Writing agents across nine angles (brand voice, third-party voice and word of mouth) apply frameworks and styles from a library of over a thousand proven formats and write from your brand knowledge base, producing social posts, long-form articles and short-video scripts for every channel. No more running out of content.
- 01 · Topic library
2,486 topics
Photo to full setIn orderRandomTrendingFinally, a Chinese-style vest I can wear every day: groceries, school run, no problem
- 02 · Writing agents, 9 angles
Short postVideo scriptLong-form Brand 3rd party Customer
- 03 · Five-step pipelineAudienceScenarioDraftPolish & complianceFinal + images
- 04 · Today's drafts · to review
27 pieces
Social post · Finally, a Chinese-style vest I can wear every day!
What New Media Content Factory is
Topics, writing, images and compliance, on one production line.
Why we built it
Every company knows it needs content marketing.
The hard part is publishing every day.
Working with companies, we kept seeing the same thing. Every channel needs regular updates, the content team is two or three people, and by the second month they run out of topics. New hires do not know the product, so the writing does not sound like the brand. Generic AI mixes up facts, reads like an ad and makes every post look the same. When the content stops, traffic and enquiries drop with it.
- Not enough content; accounts go quiet
Many channels, high frequency, and people cannot keep up. Once updates stop, platform reach and followers fall, and earlier investment is wasted.
- Same-looking content that does not convert
The same few angles over and over. Readers tune out, platforms stop pushing it, and plenty of content brings no more enquiries or sales.
- Running out of topics
Deciding what to write each day takes the most time and depends on one person's inspiration.
- Hard to hire and keep writers
People who know the product and can write are rare, and when they leave, the style and know-how leave too.
- Uneven quality
Different people at different times write at different levels, and product facts and selling points are often wrong.
- Compliance risk
Superlatives, efficacy claims and competitor comparisons easily break advertising law or platform rules.
So we turned content production into an automated line
A topic library supplies topics, writing agents across nine angles share the writing, and the brand knowledge base checks every fact. The team only reviews, and content goes out every day.
The process
From topic library to finished piece,
every step runs on its own.
- 01Build the knowledge base
- 02Generate topics
- 03Pick topics
- 04Write across 9 angles
- 05Review and publish
- 016types of brand material
Build the brand knowledge base
AutomaticBefore starting, we organize your material with you into six types (D01–D06), then keep it updated as products and campaigns change.
SmartWhich facts each piece uses, what voice it takes and which words it avoids all come from the knowledge base. Nothing is made up.
- 022,486topics · sample brand
Generate the topic library
AutomaticSelling points, target customers, use scenarios and emotions are combined into thousands of topics.
SmartEach topic states who it is for, in what situation, what problem it solves and which selling point it leads with.
- 034ways to pick topics
Pick topics in four ways
AutomaticTopics are pulled by product photo, in order, at random or by trend, and scheduled every day.
SmartFrom a product photo, AI recognizes the style and selling points and derives topics; the trending mode matches the day's holidays, seasons and platform trends to topics that fit the brand.
- 049content angles
Agents write across nine angles
AutomaticEach topic goes to the writing agent for its angle, and all nine work at the same time.
SmartEvery piece follows five steps: anchor the audience → derive the scenario → draft → polish and check compliance → finalize with images, using frameworks and styles from the style library that fit the account.
- 0536pieces / day · example
Review and publish
AutomaticDrafts wait for review; once approved, titles and formats are adapted to each platform. Drafts not handled within 48 hours are approved automatically.
SmartUsed topics are marked so the same angle is not repeated, and rejection notes are carried into the next piece.
Demo
The library feeds topics,
nine writing agents start at once.
An interactive animation of the New Media Content Factory workspace, rebuilt from the real system. All four topic modes use the same sample apparel brand and play in turn; click the tabs above to switch.
Photo to full set · by product
One product photo,
a whole set of content.
When new stock arrives or a sample is ready, an employee takes a photo and hands it to New Media Content Factory. AI recognizes the style, fabric and selling points, works out the best audience and topic, writes copy for each platform and generates images. The images use the photo as a reference, so the product in them is the real product.
01Upload a photo
02What AI recognizes and derives
- Best-fit audience
- Women aged 41–48 with a taste for cultural style, mid to plus sizes, who want something with character that still works every day
- Target scenario
- Between work and semi-formal: school open days, afternoon tea with friends, gatherings that call for neat but not dressy
- Emotional angle
- Western dresses feel wrong, older styles feel dated, a full traditional skirt is too formal: finally a Chinese-style piece for every day
- Chosen topic
- Finally, a Chinese-style vest I can actually wear every day: groceries, school run, the subway, no problem
03The full set
- XiaohongshuFinally, a Chinese-style vest I can wear every day! Groceries and school run, no problem
- DouyinCan't wear new Chinese style out? Pair this vest with jeans and go
- WeChat Channels19 years in apparel, and this is the Chinese-style vest I'd pick: if you can wear it every day, it's truly good-looking
Show the full post (translated)
After almost twenty years in apparel, the question I hear most these days is: "New Chinese style looks great, but can you really wear it out on a normal day?"
I get it. Plenty of wardrobes hold one or two Chinese-style pieces: a mandarin collar, knot buttons, embroidery. They catch your eye in the shop, then sit in the closet. Either too formal, like you're off to a banquet, or the cut is off and it hangs loose.
That's why I want to talk about this navy Chinese-style vest. Mandarin collar, a small spray of plum blossom embroidered on the left, knot-button closure. It is clearly Chinese in style. The clever part: it's a vest. A vest is an everyday shape by nature, easy over a white shirt or a light knit, and the deep navy goes with everything.
The fabric has a linen feel and holds its shape, so the shoulders don't sag. Most practical of all: you really can wear it every day. Groceries, the school run, the subway. Add a pair of jeans and you're out the door without putting together a whole outfit.
The same photo can also produce private-channel posts, video scripts, long-form articles and e-commerce product pages. This example comes from a real demo for a Chinese apparel brand, with brand details removed. New Media Content Factory writes in your market's language.
Brand knowledge base
Whether the content is right
depends on the brand knowledge base.
The knowledge base is the foundation of New Media Content Factory. Before starting, Starbox works with you to organize brand material into six types. After that, every product fact, customer scenario, tone of voice and banned word in every piece comes only from here. General rules, agents and style samples are shared, so a new brand only needs a new knowledge base.
Every selling point, with the pain point and experience it maps to, marked as must-mention or optional.
Decision types, personas, scenarios, emotions and how to respond; the main source for the topic library.
Hard rules, brand persona and style, and general writing principles.
Banned words and limits on competitors, prices, categories and emotional tone.
Real stories of the founder and the brand, built when needed.
Business knowledge, processes, methods, technical product detail and industry background.
- Facts only from the knowledge basePrices, ingredients, processes and certifications are never made up or exaggerated.
- Constraints checked on every pieceBanned words, competitor and price rules are scanned during polishing.
- Updates apply right awayWhen a product launches or a campaign changes, update the knowledge base and the next piece uses it.
Topic library · four ways to pick
One large topic library,
four ways to use it.
The topic library is generated from the selling points, target customers, use scenarios and emotions in the knowledge base. It holds thousands of topics and grows with the knowledge base. Each day topics are pulled in the four ways below and handed to the writing agents. The most used is the first: one product photo is enough.
Photo · by product
Snap a product photo; AI recognizes style, fabric and selling points, derives topics and writes a full set. Or name a product to focus output on it.
One vest photo → posts for three platforms + cover, scene and detail imagesIn order
Write down the library one topic at a time: full coverage, nothing missed, nothing repeated.
No. 128 "How long should a vest be for a pear-shaped figure" → No. 129 "First new Chinese-style purchase: which piece is safest"Random
Random combinations of customer, scenario, selling point and emotion, so even one product gets new angles.
Frequent traveler × light luggage × one piece, many looks → "One vest for a week of business travel: five outfits"Trending
Holidays, seasons and platform trends are checked daily and matched to topics that fit the brand.
First frost → "Layering for first frost: one vest, warm without the bulk"Writing agents across 9 angles
Brand, third-party, customer: nine angles, each with its own agent.
A brand needs to speak for itself, needs creators and experts to speak about it, and needs customers to speak about it. Three voices times three formats (short posts, video scripts, long-form) make nine content angles, and each one has a dedicated writing agent.
New Media Content Factory · 9-angle writing agentsVoice × format| Short post | Video script | Long-form | |
|---|---|---|---|
| Brand | A1Product posts / private channel | B1Brand video / explainer | C1Brand / knowledge article |
| 3rd party | A2Creator posts | B2Creator video | C2Expert visit |
| Customer | A3User posts | B3Customer case video | C3Case story |
Short posts: social feeds and image posts · Video scripts: talking-head, visit and case videos · Long-form: blogs, articles, press releases
Style library
Over a thousand proven frameworks and styles:
the same content, told by a different person.
Starbox keeps collecting high-performing posts from each platform, breaks down how creators talk and how successful content is structured, and adds them to the style library. After writing, the agent applies any style that fits the account: video scripts imitate a person's voice, image posts apply a structure.
Decides how the content is laid out
Topic, structure and order of information decide what a piece says and in what logic. A reusable content frame.
Decides how this person talks
Feel, wording, rhythm, catchphrases and endings decide who tells the same content. The two layers combine freely.
Imitate, do not analyze
The moment you summarize a "list of language features", imitating a real person turns into filling a template. Keeping that human feel is what makes the library valuable.
All names are rewritten English names; the demo text is rewritten from style library samples.
Results
The same team,
from "can't keep up" to "publishing every day".
Using the sample brand: a two-person content team used to publish two or three pieces a day. With New Media Content Factory, the team only reviews, and every channel is updated daily.
336
29
202,486
61%92%
Illustrative data for the feature demo.
How to use it
Use New Media Content Factory on its own, or inside Starbox Starbox!.
Deploy New Media Content Factory on its own
For companies that run content and social marketing: it writes content for every channel each day from your brand knowledge base, you review and publish, and you never run short of content.
Get a New Media Content Factory plan → Inside Starbox Starbox!Drafts on request, in one sentence
An employee says "Starbox Starbox! Write a social post" and the system hands it to the agent for that angle; you can also schedule daily production.
About Starbox Starbox! →New Media Content Factory and Starbox GEO are two separate systems: New Media Content Factory solves the content shortage; Starbox GEO is dedicated to AI search optimization.
FAQ
New Media Content Factory FAQ
What is the difference between New Media Content Factory and writing directly with AI?
Writing directly with AI requires you to come up with topics every time, produces inconsistent results, and lacks an understanding of your brand. New Media Content Factory starts by building a brand knowledge base, then automatically pulls topics from an extensive topic library across four dimensions. These are handed off to drafting AI agents across nine dimensions covering brand voice, third-party voice, and word of mouth, following a fixed five-step pipeline. Factual claims are sourced strictly from the knowledge base, ensuring output that is consistent, compliant, and non-repetitive.
What types of content can New Media Content Factory produce?
Nine dimensions: brand voice (brand product seeding and Moments posts, brand spoken videos and industry primers, brand long-form articles and in-depth knowledge pieces), third-party voice (influencer seeding, influencer videos, expert visits), and word of mouth (organic user posts, customer case study videos, case stories), covering platforms such as Xiaohongshu (RED), WeChat Official Accounts, Douyin, WeChat Channels, Moments, and Zhihu.
Where do topics come from? Will content repeat itself?
The topic library is generated by combining product selling points, target audiences, use cases, and emotional hooks from your brand knowledge base—producing thousands of topics that continuously expand. Topics are pulled daily through four modes: sequential, random, trending, and product-focused. Used topics are automatically tagged so the same angle is never repeated.
Do we need to redevelop everything if we switch to a different brand?
No. General rules, AI agents, and tone samples are fully reusable. Switching brands only requires updating the brand knowledge base D01–D06.
What problem does New Media Content Factory solve?
The problem of enterprise marketing content shortages and inconsistent output: with so many platforms and high publishing frequencies, human writers cannot keep up; relying on inspiration leads to creative burnout; and employee turnover drains brand tone and accumulated experience. The New Media Content Factory turns content creation into an automated pipeline, leaving your team to simply review and approve.
When should each of the four topic-generation methods be used?
Sequential mode is ideal for daily high-volume output, ensuring complete, thorough coverage without omissions; random mode helps maintain variety, delivering fresh angles even for the same product; trending mode taps into current buzz by automatically matching daily holidays, solar terms, and trending topics on platforms; product-focused mode suits new product launches and flagship lines that require concentrated content releases.
What do brand voice, third-party voice, and word of mouth mean?
Three distinct voices based on 'who is speaking': brand voice represents the brand owner and core team to build brand awareness; third-party voice comes from influencers and outside experts to build trust from an external perspective; word of mouth features everyday consumers and actual customers sharing reviews and case studies. Multiplying these three voices across three formats—short graphic posts, video scripts, and long-form articles—creates nine distinct content dimensions.
What materials do we need to prepare for the brand knowledge base?
Six categories: D01 Product Arsenal, D02 Target Audience and Scenario Library, D03 Writing Rules, D04 Brand-Specific Constraints, D05 IP Story Library, and D06 Enterprise Knowledge Base. Starbox works with you to compile them; you only need to provide product information, customer profiles, and existing top-performing content.
How much content can be produced each day?
Production volume is configured based on your platform needs and publishing frequency. An example brand produces around 36 pieces per day consistently, covering Xiaohongshu (RED), WeChat Official Accounts, Douyin, WeChat Channels, Moments, and other platforms (illustrative data).
How do you ensure content quality and compliance?
Facts are drawn strictly from your brand knowledge base with zero hallucination. Every piece follows a five-step process: anchoring the audience, developing scenarios, generating drafts, polishing and compliance checking, and final output with imagery. During polishing, every draft is scanned for advertising law superlatives, brand-restricted terms, competitor mentions, and pricing constraints before human review.
Does our team need to review the content?
Yes. Completed drafts enter a review queue where you can approve, reject with feedback, or download them. Items left unreviewed for over 48 hours are automatically approved. Feedback from rejected drafts is incorporated into subsequent writing.
What kind of content can be generated from photos?
An employee takes a product photo, and the AI identifies the style, fabric, and selling points, determines the ideal audience and topics, and generates titles and copy for Xiaohongshu (RED), Douyin, WeChat Channels, and more. Using that photo as a reference, it generates cover images, lifestyle scenes, and detail shots showing the actual product. The same photo can also generate WeChat Moments posts, spoken video scripts, WeChat official account articles, and e-commerce product detail pages.
What is the tone library? Can we use our own brand tone?
The tone library is New Media Content Factory's built-in repository of viral frameworks and writing styles, featuring 895 spoken-script tones, over a thousand viral graphic-text frameworks, and more than 35,000 style samples. When drafting content, you can apply any style to match your account's positioning: spoken scripts emulate distinct personal voices, while graphic posts follow proven structural formulas. Top-performing content from your own accounts can also be organized into your knowledge base as proprietary tone samples.
What is the difference between New Media Content Factory and Starbox GEO?
They are two separate systems. New Media Content Factory solves social media content shortages for daily operations across Xiaohongshu (RED), WeChat Official Accounts, short videos, and other platforms. Starbox GEO focuses specifically on AI search optimization, aiming to get AI to recommend your brand when buyers ask questions.
Guides
Methods and field notes
AI Content Marketing
How to Audit Your Zhihu Crisis Knowledge Base
Calm Zhihu PR responses rely on documented product trade-offs and operational limits, not defensive boilerplate. Audit your AI knowledge base with this checklist.
Oct 29, 2026 · 5 min read
AI Content Marketing
How to Audit AI Crisis PR Drafts When Facing Zhihu Backlash
Zhihu PR crisis management requires technical evidence chains over emotional apologies. Enforce a strict raw-document audit before publishing.
Oct 29, 2026 · 7 min read
AI Content Marketing
What Does Reviewing AI Content for Douyin Actually Cost?
Reviewing every AI script manually costs more than drafting it. A tiered audit framework cuts compliance review costs while maintaining account safety.
Oct 28, 2026 · 5 min read
AI Content Marketing
Should You Switch from Daily Posting to Wave-Based Distribution?
When marketing budgets shrink, replacing daily flat posting with wave-based AI distribution concentrates reach and makes incremental ROI measurable.
Oct 28, 2026 · 6 min read
AI Content Marketing
Why 1,000 AI Videos on Douyin Brought You Zero Sales
Evaluating AI video teams on output volume rewards low-quality spam. For new product launches on Douyin, KPIs must track retention and qualified leads.
Oct 27, 2026 · 4 min read
AI Content Marketing
Standardizing AI Visuals for Douyin Product Launches
Eliminate mismatched AI product graphics on Douyin by locking reference anchors, modularizing background prompts, and separating text from image rendering.
Oct 27, 2026 · 5 min read