AI text generation
In the new post composer (“Content” → “New post”) there’s a chat with the content agent on the left. You describe in words what the post should be about — and the AI returns several ready-made text variants with hashtags that apply to the editor on the right in one click. No need to fill out forms: a plain phrase like “announce a 20% discount on coffee this week, friendly tone” is enough.
You can see which agent writes the texts in the chat header. If you have several content agents, you can switch between them there too, and tap the pencil icon to go to the agent’s settings (model, prompt, databases). More on this in the AI agents section.
How to ask for a post in the chat
- Open “New post” and pick the platforms to publish to (chips on the chat’s welcome screen). If you don’t pick any, generation will use all connected social networks.
- Write a brief in the field at the bottom of the chat: topic, occasion, key details, desired tone.
- The agent returns several text variants — as cards with character counts and hashtags.
- Tap “Use” on the variant you like — the text drops into the editor on the right.
Random ideas
Don’t know what to write about? In an empty chat, 4 random post ideas are shown. Tapping an idea immediately sends a request to the agent — no extra steps. The 🎲 button reshuffles the selection and shows new ideas.
If no platforms are selected, the idea will be generated for all connected social networks — the choice will be reflected in the platform picker.
”Full post” — text and images in one tap
The “Full post” button in the quick-actions row launches everything at once: the agent writes text variants, and in parallel 3 image variants are generated for the post. You pick the best text and the best image — and the post is ready. For image generation settings, see the Media generation section.
”From link” — a post from a URL
The “From link” quick action: paste a page address — a product card, an article, a landing page. Repost21 will extract the page content itself (title, description, text), build a brief from it, and run the same full-post scenario — text plus images. Handy for product announcements and reworking articles for social networks.
Refining a variant with chips
Under each generated variant there are one-tap refinement chips:
| Chip | What it does |
|---|---|
| Shorter | Trims the text while keeping the gist |
| Longer | Expands the idea in more detail |
| ➕ emoji | Adds fitting emoji |
| Different tone | Rewrites it in another tone of voice |
| 🔁 More | Generates a new variant on the same topic |
Each chip sends the variant to the agent together with an instruction — and a reworked version comes back in response. You can refine as many times as you like.
Applying to the editor
- “Use” on a variant card — the text and hashtags drop into the active editor tab.
- “Apply all variants” — if texts were generated for different platforms, each one goes into its own tab.
- In the editor you can add to the text by hand, add emoji, media and hashtags — the AI variant is just a starting point. From there it’s a draft, publishing right away or on a schedule: see Posts and content plan.
Post language
Repost21 separates the interface language from the content language: the language of your menus doesn’t affect the language the AI writes posts in. The output language is configured separately, on three levels:
| Level | Where it’s set | When it applies |
|---|---|---|
| Post language | The “Post language” picker in the composer — Auto / Russian / English / Oʻzbekcha / Español | For a specific generation, overrides everything else |
| Brand content language | The “Team” → “Brand content language” section | Default for all of the team’s generations, see Team |
| Per-platform language | A toggle in the composer, appears when 2+ platforms are selected | One post in different languages for different audiences |
How it resolves: if a post language is set, it’s used; otherwise the brand content language; if nothing is set anywhere (“Auto”), the AI writes in the language of your brief.
Per-platform language — for cases where the audiences differ: for example, a Telegram channel in Uzbek and LinkedIn in English. Turn on the toggle, assign a language to each platform — the agent will generate variants for each language group, and the text in its own language will go into each editor tab.
Brand context — style and facts automatically
If the content agent has brand resources attached, they’re mixed into every generation on their own, without reminders in the brief:
- Brand library — brand voice, signature colors and style: the texts sound “like you”, and the generated images follow your signature style.
- Knowledge base — facts about the product, prices, terms: the agent relies on real data instead of making things up.
You can attach a knowledge base and brand library in the agent’s settings — see AI agents. Additionally, the brand library for the current post’s media generations can be switched right in the composer header.