Writing posts
Social studios share one editor: brief the post, draft it with AI or write it yourself, then edit, version, illustrate and check it before publishing.
Watch the video
Every network in Social has a post studio built the same way: one form that holds the brief, the copy, the pictures and the publishing of one post. This article is for editors, and it covers the parts every studio shares. The network articles (LinkedIn, Instagram, Facebook, TikTok and X) cover what’s different on each.
The short version
- Open the network under Social and check the brand in the header.
- Write the brief and click Draft with AI, or click Write it myself.
- Edit the copy. It saves itself as you type.
- Add a picture or a video on the pictures step, if the post needs one.
- Check the voice, send it for validation if needed, then publish.
The rest of this page goes through each step in detail.
Open a studio
Pick a network under Social in the left menu: LinkedIn, Instagram, Facebook, TikTok, X, WeChat, Xiaohongshu, Douyin or Weibo. Make sure the header is on the right brand first.
The dark band at the top of the studio holds:
- the post list button, which shows the open post’s title (or All posts) and how many posts the brand has
- How this page works, a short note about the network
- New post, for editors, which clears the form for a fresh post
- the steps of the post as tiles. On an open post each tile is a tab. The tile shows the step’s state: done, in progress, to do or not needed.
On LinkedIn, Facebook, X, WeChat, Xiaohongshu, Douyin and Weibo the steps are 00 The brief, 01 The copy (or 01 The title and copy when the format takes a title), 02 The pictures (or 02 The video) and 03 Publishing. On Instagram and TikTok they’re 01 the picture or the video, 02 The caption and 03 Publishing.
The copy and pictures steps show their Next button twice: in the header of the card, so the way on is in sight without scrolling, and again at the bottom. It names where it leads, such as Next: the pictures. On a text-only post it leads straight to publishing.
Find a post
Click the post list button to open a panel over the page with every post of the brand. In the No brand workspace, you only see your own posts and the ones colleagues shared.

- Type in the search box to match words in the title or copy, the status, or a date written the way you’d write it (2026-09, September, 14/09/2026).
- Use Written from and to to keep posts written between two days. Clear the dates drops them.
- Each row shows the status, the day it was written, the number of pictures, whether it’s queued and a dot per step.
Click a row to open that post. The panel closes when you click elsewhere or press Escape.
Write the brief
A LinkedIn, Facebook, X, WeChat, Xiaohongshu, Douyin or Weibo post starts with a brief. The form asks for the frame first, and the words after.
The brief takes the whole width of the page. The shape, the language and the model sit on one line above it, and the skills and the material wait in a strip at the right edge labeled Skills and material. Click the strip to unfold their two cards, which slide in from the right; click it again to fold them away. Your browser remembers whether you left it open.

- Pick the shape: Text only, + Image or + Carousel. Facebook also offers Video. On X the several-picture shape is + 2 to 4 pictures. On WeChat, Xiaohongshu, Douyin and Weibo, you pick one of the network’s native post formats instead, and the brief shows the fields that format needs (see below).
- Decide on emoticons. Emoticons is ticked by default, which asks for 3 to 6 emoji spread through the post. Untick it and the copy comes back without a single one.
- Pick the language. The brand’s own languages come first in the list. On WeChat, Xiaohongshu, Douyin and Weibo there is no language picker: the post is always written in Simplified Chinese (Written in Simplified Chinese. Brief it in English or in Chinese.), whether you write the brief in English or in Chinese.
- Pick the model. The model button opens on Quick, Balanced and Best, each with the model it runs and what one post costs on it; the card marked Recommended here is the network’s default, set by your admins (see AI models). All models lists every model your organization allows, minus the ones you switched off in My models. The app remembers your pick for next time. See choosing a model.
- Write the brief the way you’d brief a writer: the angle or the news, who it speaks to and what the post has to achieve. The box holds 20,000 characters, and the counter above it turns amber near the end.
- Leave Brand voice on to write with the brand guidelines, or its learned voice, and score the draft against them. Turn it off to write without them.
- Click Draft with AI.
On the four Chinese networks, each format card shows whether it is published from a Computer or phone or the Phone app only, unfolds What … takes, and what the draft writes with the exact limits, and asks for that format’s own brief fields:
- WeChat: Official Account article (
公众号图文, up to 20,000 characters, with author作者, length,阅读原文read-more link,原创声明original declaration and a 120-character摘要summary), Official Account image post (贴图, 1 to 20 pictures and up to 1,000 characters), Official Account short text (文字消息, up to 300 characters), Official Account video (视频消息, title and 120-character description), Channels video (视频号视频, 6 to 16-character短标题short title, up to 1,000 characters with#话题topics, location,扩展链接Official Account article link and AI declaration) and Channels image post (视频号图文, 1 to 18 pictures and up to 1,000 characters). - Xiaohongshu: Picture note (
图文笔记, 1 to 18 pictures, 20-character title, up to 1,000 characters with#话题topics, optional封面文字cover line, location, poll and AI declaration), Video note (视频笔记, up to 15 minutes, 20-character title, optional cover line,章节video chapters, location and AI declaration), Text card note (文字配图, phone app only, writes up to 140 characters of文字配图card text alongside the note title and body) and Long article (长文, phone app only, up to 6,000 characters with a 100-character摘要summary). - Douyin: Video (
视频, 30-character title, up to 1,000 characters with#话题topics, optional封面文字cover line, location and AI declaration), Picture post (图文, 1 to 30 pictures, 20-character title, optional background music mood suggestion, location and AI declaration) and Long article (文章, 30-character title, up to 8,000 characters and a 100-character摘要summary). - Weibo: Post (
微博, up to 2,000 characters folded after the first 140,#话题#closed topics, optional link, poll and giveaway via微博抽奖平台), Picture post (图片微博, 1 to 9 pictures, optional link and poll), Video post (视频微博, 6 to 30-character video title, up to 2,000 characters,原创original declaration and合集collection) and Headline article (头条文章, 32-character title, 44-character导语lead and up to 20,000 characters).
The Campaign choice next to Draft with AI files the post, its pictures and every autosaved edit in a campaign. The pictures step shows the same choice. See Campaigns.
Import a file next to the brief reads a .txt or .md file in your browser and adds its text after what’s already in the box. The file isn’t stored. Front matter is dropped and blank lines are evened out. Files over 2 MB are refused, and a file that would overflow the box is cut at the limit.
X adds its own settings to the brief (shape, hashtags, mentions, a link). See X.
Tip: Don’t need AI? Write it myself opens the editor with whatever is already in the brief box as the first text. No model is called and nothing is billed. The button only shows for a new post.
Material to write from
Material is optional, and it lives in the Skills and material strip: unfold it to add some. Whatever you add there is read once and kept with the post.
On Instagram, TikTok and YouTube there’s no brief, but the same strip sits beside the first card, where the picture or the video is made. What you add there is read when the post is created by your render, pick or upload, kept with the post with the skills you picked, and the caption (or the title and description) is written from it later. See Instagram.
| Field | What it takes |
|---|---|
| Files | Up to 5 PDF or plain-text files (.pdf, .txt, .md, .csv, .json). The text is taken out and kept, the file itself is never stored. A file that’s too large is skipped and named under Could not read. |
| Context folder from the Assets Library | Every text document in that folder and its subfolders (PDF, Word, Excel, plain text) is read before writing starts. Your campaigns are listed under the folders: pick one and its brief, a line per asset and the text of its documents are read instead. Manage the library opens it in a new tab. |
| Pages to read | Up to 5 web addresses, one per line. Each page is fetched and stripped to its text when you draft. |
| Keywords to target | Type a term and press Enter, up to 25. They’re worked into the copy as it’s drafted, and the brand’s tracked keywords are suggested as you type. |
Skills
The Skills card, in the same strip, is the skills picker you’ll find elsewhere in the app, and the network’s own format skill is already picked: LinkedIn post format, X post format, Facebook post format, Instagram caption format, TikTok caption format, YouTube title and description format, WeChat post format, Xiaohongshu note format, Douyin post format or Weibo post format. It carries the network’s posting rules.
With the LinkedIn, X, WeChat, Xiaohongshu, Douyin or Weibo format skill picked, the draft is checked against the rules after it’s written. Anything broken goes back to the model once to be fixed. Unpick the format skill to draft free-form, with no checks. You can add your own skills too, and Manage skills opens them in a new tab to read or edit. See skills.
The copy step has a skills picker of its own, beside the model under Another version (and in the Have AI write the caption box on Instagram, TikTok and YouTube). It opens on the skills the post was started or drafted with, and on the network’s format skill when none were kept. See versions.
One case has no picker at all. A post re-purposed from another network is drafted with the target network’s format skill alone. See publishing.
After you click Draft with AI
The draft opens on the copy step. A line says Drafted with what it cost, and the same line shows under Another version on the copy step. With Brand voice on, that cost includes the voice score taken right after the draft. Two warnings follow when they apply:
- Could not read: a file or page from the material that failed.
- Still off the rules: a rule the fix pass couldn’t land.
The copy is saved as version 1, a copy goes to the Assets Library as a text, and, with Brand voice on, the post gets a voice score. If the model sends back no working title, the post is named after its first line. The brief box empties, ready for the next post. The run also shows in the Activity panel, so you can leave the page while it works.
Edit the copy
The copy step has a single editor, with a working title on top and the copy under it. Whatever sits in that box is what goes out. While the post can be edited, the step is split down the middle: what asks the AI for a text sits on the left (Another version, and the Have AI write the caption box on Instagram and TikTok), and the version switch and the editor sit on the right.

- The counter under the editor shows how much room the network leaves. LinkedIn counts to 3,000 characters, X counts each post of a thread in weighted characters, and WeChat, Xiaohongshu, Douyin and Weibo count against the chosen format’s limit (with a separate counter on the title field whenever the format carries a title). The other networks show the character count.
- Add an emoticon opens an emoji picker.
- Save forces a save.
On the publishing step, you can edit the preview of the post as well. Type straight on it, and what you type goes into the same copy, with the same autosave.
Autosave
You rarely need Save. The post saves itself about a second after you stop typing, as soon as you leave the field, and when you hide or close the tab. A line next to Save says Saving…, Saved or Not saved: with the reason. If a save fails, your text stays on screen and the next keystroke tries again. Opening a post never writes anything.
Save earns its place after a big change. It saves and then reloads the post, so the counter and the version list catch up with you.
Versions
A post keeps every version of its copy, numbered v1, v2 and so on. Once there are two or more, a Versions switch appears above the editor. The green dot marks the current version: the one the network, the library and validation get.
- The line above the editor describes the version on screen: its number, the day it was written, the instruction it was asked for and, when AI wrote it, what it cost.
- Click a version to load it into the editor.
- Use this version makes the one on screen current.
- Delete removes the version on screen. The others keep their numbers. It only shows when the post has more than one version.
When the wording is close but not there, type what to change under Another version (“shorter”, “more technical”, “end on a question”), pick a model and click Write another version. The Skills dropdown on the same line holds the network’s format skill by default. Add your own skills or unpick it, and Manage skills opens them in a new tab; the post keeps your choice while you work on it. The rest of the post is kept, the new version becomes current, and a line under the box gives its cost.
Rewrite one passage
Highlight a passage in the editor or in the preview, and a Rewrite with AI button appears next to it. Right-clicking the passage opens the same menu. Pick a quick edit (Shorter, Longer, Plainer, More concrete, Fix the writing, More emoticons or Fewer emoticons) or type your own instruction. Only that passage is rewritten, and the result is saved as a new version.
Draft again from the brief
On LinkedIn, Facebook, X, WeChat, Xiaohongshu, Douyin and Weibo the brief stays a step of the post. Click the 00 The brief tile and the boxes come back filled with what the post was written from, including the native format and its fields on the Chinese networks. Change what needs changing, say a sharper angle or another format, and click Draft again. The copy that comes back is a new version of the same post, and the old one is kept. That’s the move to make when the angle itself was off.
Files, pages and folders aren’t put back in the boxes, because only their text was kept. If you attach nothing new, the text read the first time is used again.
Simplified Chinese, English reading and extra fields
On WeChat, Xiaohongshu, Douyin and Weibo, every post is written in Simplified Chinese, and the copy step adds three panels so non-Chinese speakers and validators can work with full confidence:
- The other fields of the … appears right under the editor whenever the chosen format writes extra composer fields beyond the title and body: a Summary (摘要) on a WeChat Official Account article, Xiaohongshu long article or Douyin long article, a Lead (导语) on a Weibo headline article, a Cover line (封面文字) on a Xiaohongshu or Douyin post, Card text (文字配图) on a Xiaohongshu text card note, Chapters (章节) on a Xiaohongshu video note, Music to look for on a Douyin picture post, or Poll question and Poll options when a poll is enabled. Each field shows its character counter, a Copy button, its English translation underneath and Save these fields.
- What it says, in English sits below the editor and shows a faithful, line-for-line English reading of the Chinese title, copy and extra fields, with
#话题topics explained in brackets. It also flags anything the network would refuse or penalize: off-platform links, phone numbers or WeChat IDs on Xiaohongshu and Douyin, unclosed#话题#tags on Weibo, or absolute claims forbidden by China’s Advertising Law (广告法, such as最佳,第一or顶级). If you edit the Chinese text directly, a notice warns that the reading is from an earlier wording, and Read it again in English (or Read it in English) translates the current Chinese back into English. If you read Chinese yourself, tick I read Chinese and the English reading stays folded in your browser. - Write it in English, publish it in Chinese, on the left side of the copy step, lets you write or paste a title and post in English (or any language you read), pick a model and click Translate into Chinese. The app localizes it into native Simplified Chinese for that network and format, saves it as the next version of the post with its faithful English reading beside it, and updates the format’s extra fields.
Add pictures or a video
Once the words are settled, the post may still need something to look at. The pictures step starts with the shape. You can change it at any time (on the Chinese networks, the format picked on the brief sets whether the post takes pictures or a video), and it’s saved as soon as you pick it. Pictures already on the post stay until you take them off.
The rest of the step is split in two. On the left are the ways to add a visual and the render settings. On the right is what’s on the post, which stays in sight while a render runs, with the picture versions under it.

Then pick one of three ways in:
- Render with AI (Render again once there’s a picture). The price sits on the button: per image, per render for a carousel, or for the chosen seconds of a clip.
- Pick from the library (Add from the library on a carousel that already has pictures). Use a picture or clip already in the Assets Library.
- Upload from your computer. The file goes into the Assets Library under the post’s brand and onto the post. Pictures can be PNG, JPEG, WebP, AVIF or TIFF, plus GIF on LinkedIn, X and Facebook (Instagram and TikTok take no GIF), and a carousel takes several at once. Where a network doesn’t take a picture as it is, the app converts it to a JPEG when it sends the post, and the copy in the Assets Library doesn’t change. A clip has to be a type the network publishes, and it’s checked before a byte is uploaded.
Under the three buttons sit the render settings:
- Created with AI: the prompt. Describe what the picture shows. For a carousel, write one block per slide (“Slide 1: headline”, then the prompt). Left empty, the prompt is written from the brief and the post when you render. On LinkedIn and Facebook, the first time this step opens on a post that has copy but no picture yet, the box fills itself from the post, as Improve with AI would, with the cost in the note under it. That happens once per post, and whatever you type in the meantime wins.
- Improve with AI writes the prompt for you from the brief and the post. It’s billed, and the cost shows next to it. The model picker beside the button chooses the text model that writes the prompt. Edit the result freely before rendering.
- Engine and Aspect. Each engine shows its price in the list, and the list opens on the cheapest engine, so a render you start without touching it costs the least.
- For a clip, Length, Resolution and With sound, when the engine offers them.
- For a carousel, Slides: 2 to 8, or 2 to 4 pictures on X.
The settings end with a button of their own, priced like Render with AI and doing the same thing: Generate the image, Generate the carousel, Generate the pictures on X, or Generate the clip. Once the post has a visual, it reads Generate a new image (or a new carousel, new pictures, a new clip).
Expect a minute or two for a picture. A clip takes a few minutes, so feel free to leave the page: the run shows in Activity and the post keeps the result. When it’s done, the line under the buttons says Done with the cost.
A single picture replaces the one on the post. On a carousel, new pictures are added at the end, up to 8 (4 on X; on a Chinese format that accepts more than 8 pictures, render up to 8 and add the rest from the library or your computer). Click a picture to open it large. The × on a picture takes it off the post, and it stays in the Assets Library. The pencil opens the picture in the Image editor, on its Social panel for an X, Instagram, LinkedIn, WeChat, Xiaohongshu, Douyin or Weibo post: Save and use it in the post puts the edited copy in its place, on the same slide, as a new picture version, and the original stays in the versions and in the Assets Library. On a video post, the pencil on the clip opens the Video editor the same way. The pencil is easy to miss, so a button under the pictures does the same: Edit in the image editor (Edit slide 1 in the image editor on a carousel, where every slide keeps its own pencil) or Edit in the video editor.
Picture versions
The copy has versions, and so do the pictures. Every set of pictures the post has carried is kept, numbered v1, v2 and so on: each render, edit, pick from the library and upload makes a new one. Once there are two, Picture versions lists them under the pictures, newest first, each with a thumbnail, how many pictures it holds and when it was made. The one on the post is marked On the post.
- Use this version puts that set back on the post, and the line under the buttons says This version is on the post again.
- The × on a version takes it off the list. Its files stay in the Assets Library, and the version on the post can’t be removed.
Nothing here deletes a file: a picture you no longer use stays in the Assets Library until someone deletes it there.
On Instagram and TikTok this step comes first, before any copy exists. See Instagram and TikTok.
A listen version of the post
Every post, on every network, can be read aloud: the Listen version box sits at the foot of the pictures step. It reads the post, or the article the post was written from, in the voice of the brand’s sonic brand kit (when the text is in another language than the kit’s, a native speaker of that language, of the same gender, reads it), and files the MP3 in the Assets Library next to the post. On LinkedIn and X it also makes an audiogram, the listen version as a square or upright video that becomes the post’s video: see LinkedIn and X.
Checks and status before publishing
On the publishing step, the bar under the phone preview holds Save and two checks. Check voice scores the post against the brand guidelines, or against the learned voice when there are no guidelines. A line beside the buttons then says Voice checked with the score out of 100 and what the check cost. The score also shows on the band as voice and a number, green from 80, amber from 60 and red below, and hovering it shows the most useful fix.
Send for validation hands the post to a colleague to sign off. Until they decide, the post is locked, and nothing on it can be changed or deleted. The validation article covers the other side of the request.
The rest of the step is covered in publishing: the two publishing tabs, the schedule, the queue, Where it stands for setting the status and the published address by hand, and Re-purpose for other networks, which turns this post into posts for the other networks.
Share or delete a post
A post in a brand is seen by everyone in the organization. A post in the No brand workspace is its author’s alone until they click Who sees it on the band and pick Only me or Everyone in the organization.
Delete on the band removes the post for good, after a confirmation. Admins can delete any post. Authors can delete their own No brand post while it’s still Only me. Nobody can delete a post while it’s waiting for validation.
Other ways to start a post
- From Feeds, now a module of its own in the left menu, Create from this post on an article opens a studio with a brief written from what you were reading. See feeds.
- A link from the calendar, the Assets Library or an email opens a post directly, on the right step.
Who can do what
Viewers can open the studios and read posts. Briefing, drafting, editing, rendering and sending for validation need the editor role or higher. Your admin can adjust the Social Content area for you on your user page, as roles and rights explains. See brands.
What it costs
These actions are billed, and each one shows its cost where it ran, in your organization’s currencies: Draft with AI, Draft again, Write another version, a rewritten passage, Translate into Chinese, Read it in English (or Read it again in English), Improve with AI, a render and Check voice. The voice score taken right after a draft is part of that draft’s cost. Every charge is also listed in the usage log. When the LinkedIn, X or Chinese format check sends a draft back for fixing, the price shown covers both passes. What a draft costs depends on the model and on how much it reads and writes. The model picker shows what one run costs on each model before you pick. Write it myself and typing in the editor are free. Uploaded files are stored in the Assets Library and count toward storage. See what costs credits.
Troubleshooting
“Write the brief first.” The brief box is empty. Write at least one line on the angle, the audience and the goal.
“The draft ran past the length the model can answer in and came back cut off.” Shorten the brief, or ask for a shorter post, and try again.
“No draft came back.” Rephrase the brief and try again.
“Nothing to score against yet.” The brand has no guidelines and no learned voice. Write the guidelines in brand foundation first.
“Not saved:” next to Save. The autosave failed, and the reason follows. Your text is still on screen, and the next keystroke tries again. If it keeps failing, copy your text somewhere safe and reload.
“The draft came back over … limit twice.” The LinkedIn or X check couldn’t bring the draft under the network’s hard limit. Tighten the brief and try again. See LinkedIn or X.
“Pictures only: pick a PNG, JPEG, WebP, GIF or AVIF file.” You uploaded something that isn’t a picture on a picture post. On Instagram and TikTok the message names no GIF. Switch the shape to a video if the network offers one, or pick an image.
“This network cannot publish one of these files: pick a PNG, JPEG, WebP, GIF or AVIF file.” One of the pictures you picked, from your computer or the library, is a type this network can’t receive and the app can’t convert. On Instagram and TikTok the message reads “This network publishes no GIF, nor one of these files: pick a PNG, JPEG, WebP or AVIF file.” Pick another picture.
Nothing can be edited. The post is waiting for validation. Open Validation from the note on the band.