
Project info
This is where your brand's personality lives. The more precise the settings, the better the posts.
1Post language and audience region
Right in the project-name block you set the post language and audience region — these are PROJECT-level settings:
- 🗓️ all posts are written in the chosen language;
- 🌍 the region helps adapt the text culturally;
- ✨ the «Improve project» feature phrases its suggestions in the project's post language (not the language of current values), so context and plan don't contradict each other.
2Tone of voice
Describe the style, audience and forbidden topics — the AI writes accordingly.
3Marketing context
Key topics used to find and select content.
4Extra context
Instructions for the model:
- 📝 channel signature and source links;
- 🧠 a lively human style without AI clichés;
- ⛔ by default — no pushy advertising, violence or extremely negative content.
The «↺ Reset to recommended» button restores this safe text.
5Editorial rules («Note for the future»)
Every post in the queue has a «✍️ Note for the future» field: type a one-off remark (e.g. «don't advertise competitors», «more specifics, less fluff») and the model distills it into a short permanent project RULE. After that:
- 📌 rules are stored in «Project info» and applied to ALL future generations (they take priority over other instructions);
- 🧠 this is how the system adapts the context to you and stops repeating the same slips;
- ✏️ you can edit the rules list manually in project settings (one rule per line).
This is separate from the built-in «human writing style» that already applies to all posts (see «How generation works»).
6Context files (.md and .pdf)
Under «Marketing context» and «Extra context» you can upload one or more .md and .pdf files and delete the ones you don't need:
- 📎 a brand book, price list, style guide or product descriptions all fit;
- 📝 text is extracted on upload (automatically for PDF) and added to the matching context block on every generation — posts take your materials into account;
- ⏱️ limit — up to 5 MB per file; PDF pages that are images only, with no text layer, may not be recognised.
7Improve and optimize the project
The «✨ Improve» button works step by step:
- In a popup it asks what the channel/posts should be like — describe it in your own words or pick a hint (news, entertainment, corporate, advertising, educational, personal blog). The model adjusts the tone, topics and rules accordingly.
- It sends the settings to the model for review as generation context and either asks clarifying questions (about the company, product, channel, audience) or proposes rewritten field values — clearer and more useful, without fluff or AI clichés.
- Besides meaning, it OPTIMIZES the context for tokens: makes fields more compact, removes repetition and duplication between fields — the model still gets enough context, but each request costs less (the context goes into EVERY request).
- Accept each change (it fills the form) or reject it, then «Save changes».
- After saving, a brief summary lands in «Notifications»: which fields were compacted and roughly how many tokens the context shrank by (i.e. how much you save on every generation).
8Timezone and quiet hours (savings)
At the bottom of «Project info» you can set the project timezone and «quiet hours» — an interval (e.g. 23:00–07:00) when the scheduler does NOT sweep sources or generate posts:
- ⏱️ direct token savings: at night the feed usually isn't needed, while background search/generation would still cost money;
- 🗓️ time is computed in the project timezone (empty by default = server time; set your zone with the «From browser» button);
- ✅ the manual «Generate drafts» action ignores quiet hours — it always works;
- 👤 the timezone default comes from your browser and is stored in «Profile».
9Getting the most out of Improve
- Before running, fill at least the basics: name, what the brand does, who the audience is — so the model has something to work with.
- Answer the clarifying questions specifically: company and product, the audience and its pain points, the desired tone, what must not be mentioned, the channel link/topic.
- Work iteratively: answer → get edits → accept the good ones → run again if needed.
- Accept edits one by one and review the wording before saving.
- The effect is much better with a capable text model and a configured key («Profile» → keys).
10Import a project from a website
The project wizard and «Project info» have an «Import from a site» field. Enter the company website address:
- 🔍 the service reads the homepage and 3–5 relevant pages (about, products, services) and fills the project settings with the model;
- ✅ values are placed into the fields — review and save;
- ✨ the context is produced compact (no filler or repetition), so the project is marked as «optimized» — this saves tokens on every generation;
- ⚙️ if you later edit context fields manually, the mark is cleared (context changed) — you can press «Improve project» again;
- 🔑 a personal model key is required; sites that require sign-in or render fully via JavaScript may not be parsed.
11Style, models, tokens and webhooks
Here you also configure:
- ✨ the visual style — its description is added to every prompt for BOTH images AND video (when «animating»), so everything stays on-brand;
- 🤖 model selection (text / images / video);
- 🎬 auto-animation of images into video and clip duration — see «Animating images into video»;
- 🔑 API tokens for agents and webhooks.
See «AI providers & API keys» for details on providers and keys.
12Project backup
At the bottom of the page there's a «Project backup» block:
- 🧾 export all settings to JSON and restore a new project from a file;
- 🔒 secrets (bot and custom-model tokens) are not included and are re-entered after import.
See the «Project backup» help section for details.
Didn't find an answer? Write to us — we'll help.
Contact support