AI Studio

AI Studio (/app/studio) is a chat with an AI that writes posts, draws pictures, makes short videos, proposes campaign ideas and rewrites text. Every result can become a post in one click. It needs the Create task: a member without it who opens /app/studio sees "You do not have access to this" instead of the Studio.

Before you start

• AI access: your own key or included AI. Without either the screen says No AI provider is connected yet with Add an AI key in Settings. See What the AI features need.
• To publish a result you need at least one connected account.

Step by step

1. Open AI Studio in the menu. A new chat starts with "What should we create today?" and a few starter prompts. A starter puts its prompt in the box and picks its mode; anything you already typed is kept, with the starter added under it. Translate a post asks for [language] and ends with [Paste your post here]; type the language, then what you paste replaces [Paste your post here]; until it is replaced Send stays grey (the chat cannot read your posts by itself).
2. Pick a mode with the mode button on the right of the input row: Chat, Image, Video, Transcribe, Campaign ideas or Rewrite.
3. Optional: press the settings chip (it shows the current settings, such as 1:1 · ×1 · Standard) to open Options.
4. Type what you want and press Send (or Enter; Shift+Enter makes a new line; Esc stops a reply that is being written).
5. Use the result: Use in post or Make a post from this opens the Create form in a side panel, already filled in; pick accounts there and send (see Create). From a chat answer, Use in post takes only the post's words: the answer's Caption: part when it has one, with a CTA: or Call to action: part after it added as its own paragraph (otherwise the answer without its opening line, part labels such as Hook: and any closing note), without the bold and other formatting marks, and the hashtags at its end go into Hashtags. The model's closing remarks are left out too, even after the hashtags: a follow-up offer such as "Want me to write a shorter version?", "Let me know if you'd like…" or "Feel free to tweak the tone!", and a closing line all in italics; a call to action to your audience, such as "Let me know in the comments", stays. Only the end of the answer is checked, and an offer counts only when it is about the writing (it asks to write, rewrite, shorten, translate or tweak, or names a version or the tone) and asks nothing of your audience: a line to your audience such as "Want me to share the full recipe? Comment YES below", "I can also do custom orders" or "Feel free to share this post with a friend" stays in the caption wherever it sits. This works the same when the caption has no blank lines. Copy still copies the whole answer.

Modes

• Chat — ask anything or describe the post you need. Text answers have Copy, Use in post and Repurpose ▾ (Instagram caption, LinkedIn post, X thread, Reel script, 5-slide carousel, Ad copy, and Translate into [language] for each of the workspace's post languages other than your app language). Answers show bold, italic text and links formatted; a link opens in a new tab.
• Image — describe the picture. Results appear on the canvas. With a finished picture in the chat, the chip Continuing from the last image means your next message edits that picture (for example "make it September, 1299 SAR"); press Start fresh to make a new one, or Continue from the last image to go back to editing.
• Video — describe a short video, or attach an image to animate it.
• Transcribe — This is coming soon on OmniBrand's AI. On a paid plan, you can use your own OpenAI or Gemini key now, under Settings › AI providers. With a key, attach an audio file (for example MP3, M4A, WAV or WebM, up to 25 MB) or a video, optionally add a note, then send. Attaching an audio file switches to Transcribe by itself. The transcript is saved in the chat with your note (or the file's name), so it is still there when you come back. "Attach an audio or video file to transcribe." appears if nothing is attached.
• Campaign ideas — type a theme (for example "National Day offers"); you get ideas, each with suggested platforms, Copy and Use as prompt.
• Rewrite — paste the text in "Paste the text to rewrite…" and type the instruction in the main box (for example "make it shorter and more formal").

Fields and options

• 📎 attach — up to 3 files per message (image, video or audio). In Image mode an attached image is edited or remixed; in Video mode it is animated; in Chat you can ask about it (leave the box empty to have it described). An audio file is only for Transcribe: it goes straight to the transcriber and is not added to the media library. If you switch to another mode with an audio file still attached, pressing send shows "An audio file can only be transcribed. Switch the mode to Transcribe, or remove the audio file to send." and nothing is sent.
• Dropping files — drag files from your computer onto the Studio: the message box lights up with "Drop to attach" ("Up to 3 files: pictures, videos or audio."), and the files you drop are attached as with 📎. Files beyond the 3 a message takes are left out: "N of the dropped files were not attached: a message takes up to 3."
• Message length — a Chat message takes up to 8,000 characters; a prompt in the other modes, and the text to rewrite in Rewrite, up to 4,000. From 80% of the limit a counter such as 3500/4000 shows above the row of buttons; over the limit it turns red ("This message is N characters; the limit is M. Shorten it to send.") and the send button stays grey.
• Options — Format (Story / Reel 9:16, Feed 4:5, Square 1:1, Wide 16:9) for Image; for Video, only the formats the chosen model makes: Story / Reel 9:16 and Wide 16:9, plus Square 1:1 for models that make it, such as Runway from a picture (Gemini Omni does not), and never Feed 4:5; Number of images ×1 to ×4 and Quality (Standard quality or High quality; shown only for a model that renders it, OpenAI's GPT Image - Gemini and Grok Imagine have no quality setting, so it is hidden for them and their pictures count as standard) and Style preset (Any style, Promo / offer poster, Photo, Lifestyle, Cinematic, Product shot, Illustration, Flat design, 3D render) for Image; Duration for Video, with only what the chosen model makes: Gemini Omni 4, 6 or 8 s in 9:16 or 16:9, Runway 5 or 10 s, silent (a format or length the model does not have is shown as the one it will make, and the chat card says the format, length and sound made, for example "Video · 16:9 · 8 s · with sound"; for older videos, made before the card showed this, it shows no length and says "silent" only for Runway videos); under it, "Sound: this model adds its own soundtrack, and it cannot be turned off." or "No sound: this model makes silent videos."; Language hint (Detect language, Arabic, English) for Transcribe; Model for every mode; Brand voice (write in the voice of your brand profile) for every mode except Transcribe; Auto-enhance for Image ("Your words are rewritten into a fuller brief before generating. Off = sent exactly as typed."). The arrow keys move between the choices of Format, Number of images, Quality and Duration. Done closes the panel; when the panel is taller than the window it scrolls.
• Model — Auto (best for this) lets OmniBrand pick (it shows "Auto · picks (model)" when it knows the model it will use), or pick one; models on your own key are marked Your key and recommended ones Recommended. Under it you see whose credits are used (see What the AI features need); it is read again after each credit, so "N of M included credits used" counts the one just made. When no model of the mode's kind is available, the line above the row of buttons says so (for example "No image model is available yet. ...", with Open Settings › AI providers for Owners and Admins), Send stays grey, and Options shows only Model and the switches. A picture is made in exactly the Format you picked (a model that draws a nearby shape has its picture trimmed evenly at the edges). For pictures, Auto uses your package's included credits first while any are left (when that is the setting in Settings), then your own key. The same list is in Create's Generate an image window. A picture or video is never replaced by a sample picture: when no model can make it, the reason is shown with what to do (see What the AI features need).
• ✦ menu — Enhance my prompt (rewrites your prompt into a fuller one), Propose a few directions for this (Image and Video: shows "Pick a direction - you can edit it before generating"), and templates: Hook + caption, Reel script, 5-slide carousel, Product shot, Before / after, Testimonial, Offer / deal. A template puts a prompt with [brackets] for you to complete in the box and switches to the mode it is written for: Image for Product shot, Chat for the others. The first [bracket] is selected, so what you type replaces it; Tab moves to the next one. Until every [bracket] of the template is replaced, the line under the box says "Replace [product], [surface] with your own words, then send. Tab moves to the next one." and Send stays grey, so nothing is generated from the bracketed words. Anything you already typed is kept, with the template added under it; an empty box, or one still holding another template, takes the template instead.
• 🎤 Dictate (speech to text) — records your voice and types it into the box. Press Done to stop.
• ✦ Make a video from this text — under a text answer. It opens with your text, which you can edit, and a short description of what the video will show, written from the text as it stands then; you can change that too. Writing the description uses no AI credits. You choose the shape and length and see the AI credits it will use before you confirm Make the video. With a picture attached you can start from that picture. The video keeps the text it was made from, and the answer then shows those words, with See the original answer under them. Cancel leaves the answer as it was. If your own key cannot make the video, the message names that key's provider and says where to check it or choose which AI goes first. If it fails on your own key and OmniBrand's AI can make it instead, Make it with OmniBrand's AI instead appears beside the retry button, with the AI credits it will use; nothing is used until you press it. When both are available, the window also lets you choose who makes it, and your own key is selected first.
• Picture and video results (canvas) — Use in post (opens the Create panel, titled Use in post like the button, with this picture attached), Approve / Approved (marks the picture as approved; it then appears under Approved only in the media pickers), ⟳ Variations (new versions of the same request), ✎ Edit (with your own Grok key, edit this picture with a new prompt), Download, Full screen. Prompt used shows what was actually sent; a lettering chip shows Lettering checked, Text printed by OmniBrand or Lettering may be wrong for text drawn on the image.
• Chat / Canvas switch — on a phone, in a chat that has pictures or videos, shows the conversation or the large picture view. Sending a message in Chat, Transcribe, Campaign ideas or Rewrite mode switches to Chat, so you see your message and the reply; a new picture or video switches to Canvas when it arrives.
• Chat list — Chats lists your conversations with search, + New, and per chat Rename and Delete ("Delete this chat and all of its messages? This cannot be undone."). Your chats are your own: teammates in the workspace do not see them, and only you can rename or delete them. AI Studio opens on your newest chat, or on a new one when you have none. Rename opens Rename chat: a name of up to 120 characters (the counter under the field shows how many are used), then Rename or Enter. If the name cannot be saved, the window stays open with your name in the field and the reason under it. The Library tab lists everything generated, with Approved only; a video without a preview picture shows ▶. When the list slides in over the page, ✕ or Esc closes it. The list holds your 100 most recent chats; with that many, its end says "Showing your 100 most recent chats. Search to find older ones.", and typing in Search chats… also searches every older chat ("Searching all chats…"). A long chat opens on its newest 100 messages; Load earlier messages at the top of the thread shows the 100 before them. You can open another chat while a picture, video or transcript is being made: the result is saved in the chat you asked in, not the one on screen, and a notice "Your result is ready in another chat" names that chat with Open it.
• Edit & rerun on your own message runs it again after you change it; Retry repeats a failed render.

What happens next

• Every picture and video is saved in the Media library.
• When you send from the Create side panel, it closes with one message, Done, saying what happened: "Saved as a draft. Find it under Records → Drafts.", "Scheduled. See it on the calendar.", "Publishing now." or "Sent for approval." If the post reached only some accounts, the message says Sent to some accounts only and the panel stays open on the red box Some accounts did not get this post, which lists the accounts, so you can see which ones did not get it; it closes without asking, because the post is already saved.
• The side panel keeps no draft. If you changed anything in it (text, hashtags, media, accounts, an account's own version or the note) and close it with ✕, Cancel, Esc or a click outside it, a window Discard this post? opens on top of the panel with "The caption, accounts and changes you made here are not saved anywhere else." Choose Keep editing (or press Esc) to keep working, or Discard to close the panel.

If something goes wrong

• "Up to 3 attachments." — remove one first.
• "The message is longer than N characters. Shorten it and send it again." — the message went over the limit for its mode. When a Chat message is refused before it is saved, it is taken out of the thread and put back in the message box, so you can shorten it and send it again.
• "Could not load earlier messages" — the older part of the chat could not be loaded; press Load earlier messages again.
• "Could not upload the file" / "Upload failed (status)" — the file could not be stored; photos are limited to 50 MB and videos to 250 MB. With "“(file name)” is not a picture or a video ((file type)). Upload a JPEG, PNG, WebP or GIF image, or an MP4, MOV or WebM video." — the file type cannot be uploaded; this is said before anything is sent.
• "This recording is larger than 25 MB, too large to transcribe. Record a shorter one." — shorten the recording.
• "This image is larger than 8 MB, too large to look at. Attach a smaller copy." — when asking about an attached image.
• Other refusals start with a sentence in your app language (for example "Your plan's limit for this is used up." with the plan pop-up, or "No connection. Check your network and try again."). When a reply stops without a reason you see "The reply was interrupted. Please try again."
• "Render failed" under a picture or video, with a sentence that says what happened: "Something went wrong on our side while making it..." (press Retry; if it fails again, contact support), "The AI service did not accept the request..." (press Retry, or pick another Model), "The AI service took too long and stopped..." (try again in a few minutes, or ask for a shorter video), or "The AI service refused this request under its content rules..." (change the wording or the pictures, then try again).
• "Still processing — this is taking longer than usual." — the provider is slow; the result will appear in the media library when done (Check the media library).
• "Recording is not supported in this browser." or "Microphone access was denied." — allow the microphone in your browser, or type.
• "The model returned an empty answer. Try again or pick another model." Or "The AI could not answer with this model. Pick another model and try again." — the model chosen under Model cannot be used right now (for example its provider has no key here). Pick another model, or Auto (best for this).

Related

• What the AI features need
• Create
• Media library
• Brand profile