In the chat, the Elbuz AI consultant turns files and text into results: a product description from a photo, text from a scan, voice message or YouTube video, a translation, a banner, a voice-over. Attach a file with the paperclip, then type or say the task.
When you need this:
- Product photo, no description — 2–3 sentences plus the features visible in it.
- Voice message or call recording — a word-for-word transcript.
- Text to shorten, correct or translate — one button replaces the selection.
- Banner or audio — the image or mp3 goes to your files.
Where to start: open the AI consultant, attach a file, type a ready-made phrase.
What AI chat utilities can do
- Photo → description and text — a product description for the card, text from a scan, an explanation of a screenshot.
- Voice and audio → text — a voice command to the consultant or a word-for-word transcript of a call.
- YouTube → text — the text or a summary of a video, based on its subtitles.
- Summary, translation, style — also straight from the article, presentation or email editor.
- Image from description and text voice-over — the file is saved to your files.
- The same utilities in agents and flows — AI employees and the “AI agent” block have four utilities without attachments: “Text summary, translation, style”, “YouTube video text”, “Image from description”, “Text voice-over”. Voice transcription is not available there — attachments exist only in the chat. You turn them on in the agent settings → tools; details in AI employees.
How to attach a file or dictate by voice
Below the consultant's input field there are two buttons: the paperclip “Attach a photo or audio” and the microphone “Record voice (click again to stop)”.
- Images: jpg, png, gif, webp — up to 10 MB.
- Audio: mp3, m4a, ogg, wav, webm, flac, mp4 — up to 25 MB.
- The format is detected from the file contents, not from the extension.
- The paperclip adds one file at a time; click it again to add the next one.
- The file name with a cross appears above the field; while the file is uploading, you can't send the message — “The file is still uploading…”.
- Attached files are visible only to you; files older than a day are deleted when you attach the next one.
Microphone. Click it and you'll see “Recording… Click the microphone again to stop.”; click again and the recording becomes a voice message.
- Before replying, the consultant transcribes the voice into text and answers what you said; the text it heard appears in small print under your message.
- There is no duration limit, only the 25 MB file limit.
- If the browser doesn't allow microphone access — “Microphone is unavailable: allow access to it in the browser.”.
A file without text: intent buttons
If you send a file without a single word, the consultant doesn't spend an AI call. It asks “What should I do with the file?” and offers buttons:
- for an image — “Explain what is on the screenshot and help with this screen”, “Recognize the text”, “Describe the product in the photo”;
- for audio — “Do what I said” (the audio becomes a voice command) and “Transcribe word for word”.
A button does the same as typing its text yourself; “Do what I said” sends the audio itself as a voice message.
What to write: ready-made phrases
| What to attach | What to write | What you get |
|---|---|---|
| product photo | “Write a description for the product card: 2–3 sentences and a list of features visible in the photo” | description text in the chat |
| scan of an invoice, screenshot | “Recognize the text” | text from the image |
| screenshot of an error | “What's wrong here and what should I click?” | an explanation of the screen |
| voice message, call recording | “Transcribe word for word” (or no text → button) | conversation text |
| 🎤 voice | dictate: “…transcribe this and turn it into an email to the customer” | email draft |
| — | “Here's the link https://youtu.be/… — make it into text” / “— make a summary” | video text or summary |
| — | “Summarize this text: …”, “Expand on this: …” | shorter or more detailed text |
| — | “Fix the mistakes and improve the style: …”, “Rewrite in a friendly tone: …” | corrected text |
| — | “Translate into English: …” | translation |
| — | “Draw a banner: …” | image in the chat |
| — | “Voice over this text: …” | a “Listen to audio” link to the mp3 |
How to get a product description or text from a photo
The consultant reads an image with a question itself, as a picture — there is no separate “image → text” utility in the chat. Its rule: a screenshot — explain the screen and help; a product photo — answer the question; recognize text word for word or describe the product — only when that is exactly what you ask for.
AI takes features only from what is visible in the photo: color, shape, package contents. The model, battery capacity and other technical data can't be determined from an image, and AI doesn't make them up — it says that exact specifications can't be established from the photo.
How to turn voice, audio or YouTube into text
Transcription uses a speech recognition model; the language is detected automatically. There are two modes:
- voice message (microphone or “Do what I said”) — what you said becomes your message, and the consultant acts on it as if you had typed it;
- word-for-word transcript (“transcribe word for word”, “what are they saying here”, “make it into text”) — the full text of a recording, for example a customer call.
A video file (mp4, webm) is accepted as audio: its sound is transcribed.
YouTube → text
Give a youtube.com or youtu.be link — the consultant takes the video's subtitles and, by default, has AI clean them up into readable text; ask for a summary and it shortens it.
- The video's sound is not recognized: for a video without subtitles you get “Could not get the video subtitles: the video has no subtitles or YouTube did not return them. Try another video or try again later.”.
- Links to other sites — “A youtube.com or youtu.be video link is required.”.
How to shorten, translate or correct text
Shortening, expanding, fixing mistakes and improving style, changing the tone (“formal”, “friendly”) and translating are one utility. Translation languages: English, Ukrainian, Russian, Polish, German, French, Italian, Spanish; if no language is given — the interface language. The utility result the consultant sees is limited to 20,000 characters, so split very long text into parts.
Text from the editor: improve, shorten, translate
You don't need to copy the text of an article, a presentation slide or an email reply into the chat. Select it in the editor and click “AI tools” on the editor toolbar (in an email — the magic wand “AI actions”), or right-click the selection. The menu contains:
- your operations from “AI Templates” (“Correction and improvement of text style”, “Summary of the text (tldr)”, “Translate text into another language”, and so on);
- the “More tools — in the AI consultant chat” item;
- the “AI Templates” list itself.
The separate “AI assistant” window is gone: its tools are now in the consultant chat and in this menu.
1
AI tools2
Operation template3
More tools — in the AI consultant chat4
AI Templates- The button in the editor toolbar; in the email reply window it is called “AI actions”.
- Your templates: the selected text goes to the consultant chat with this operation.
- Opens the consultant chat without any text — for tasks that are not in the menu.
- The list of templates: add and edit your own operations here.
After you choose an operation, the consultant chat opens: the first line is the operation name, followed by your text (with nothing selected — the whole text of the field).
- The consultant replies with only the finished text, without introductions or explanations.
- The “Insert into editor” button under the reply replaces the selected fragment or, if nothing was selected, the whole text of the field.
- You save the article or send the email yourself.
Corrections and summaries come in the language of the text itself; in an email — in the language of the email. The translation template in an article translates into the interface language; for another language, type in the chat “translate into English: …”. Your own operations, the result language and troubleshooting are covered in “AI tools in the editor”.
How to create an image or a voice-over
“Draw a banner / generate an image / an illustration: …” — a 1024×1024 image, medium quality; description up to 2,000 characters. The image appears in the chat and is saved to your files. You can't choose the size or quality in the chat.
Text voice-over
“Voice over this text: …” — up to 4,000 characters at a time; longer text gets “Voice-over takes up to 4000 characters at a time.”. The result is an mp3 with a “Listen to audio” link, and the file is saved to your files.
1
Request to read the text aloud2
Audio link in the answer- Request to read the text aloud — a plain message asking to voice the given text.
- Audio link in the answer — click it to listen to the generated speech.
How much it costs
| Utility | Price |
|---|---|
| Image from description | charged to your balance for actual AI usage (tokens) |
| Voice and audio transcription | charged to your balance for actual AI usage (tokens) |
| Voice-over | charged to your balance for actual AI usage (tokens) |
| YouTube | getting subtitles is free; only the AI cleanup into text is paid |
| Photo → description, summary, translation, style | for actual AI tokens; the consultant reads the photo itself, so this is the cost of its reply |
| File without text → buttons | free |
All utilities are paid only with AI tokens, with no subscription. The consultant's own reply is added to the utility cost, also for actual tokens. You can see the spending in the balance charges log.
Scenarios
A supplier sent a photo without a description
What to do: attach the photo and type “Write a description for the product card: 2–3 sentences and a list of features visible in the photo”.
What you get: the description text in the chat; AI doesn't make up technical data that isn't visible in the photo (photo → description).
A customer left a voice message or you have a call recording
What to do: attach the audio and type “Transcribe word for word”, or send the file without text and click “Transcribe word for word”.
What you get: the full conversation text in the chat (voice → text).
Reply to a customer when your hands are busy
What to do: click the microphone and dictate: “…transcribe this and turn it into an email to the customer”.
What you get: an email draft in the chat.
An article is too long
What to do: select the text in the editor, then “AI tools” → “Summary of the text (tldr)”.
What you get: the shortened text in the chat; “Insert into editor” puts it in place of the selection (text from the editor).
What to do when a file is not accepted
| What you see | Why | What to do |
|---|---|---|
| “The file is too large: up to 10 MB.” / “…up to 25 MB.” | An image over 10 MB or audio over 25 MB | Compress the file or split the recording (limits) |
| “You can attach an image (jpg, png, gif, webp) or audio…” | The file is in another format, for example PDF | Take a screenshot of the page or save the audio as mp3 |
| “Attachment not found: attach the file again.” | Attached files older than a day are deleted when you attach the next one | Attach the file again (how to attach) |
| “Microphone is unavailable: allow access to it in the browser.” | The browser blocked microphone access | Allow the microphone for the site in your browser settings |
| “Could not get the video subtitles…” | The video has no subtitles | Use another video or upload the sound as a file (YouTube) |
| “The daily demo limit is used up: up to €3 a day for AI, parsing and media generation…” | The demo has spent its daily €3 limit for AI, parsing and media combined | Continue tomorrow or switch to a paid plan |
| “The account is inactive — AI features are off until the balance is topped up.” | The balance has run out | Top up your balance |
| “No access to AI utilities.” | Access to the utilities was removed from your role | Ask an administrator to restore access in the role |
FAQ
Can I attach a PDF or Excel file?
No. The consultant accepts only images (jpg, png, gif, webp up to 10 MB) and audio (mp3, m4a, ogg, wav, webm, flac, mp4 up to 25 MB).
Why can't I get text from a YouTube video?
The consultant takes the video's subtitles. No subtitles — no text: the sound of YouTube videos is not recognized.
Can I choose the image size?
No. In the chat, an image is always 1024×1024, medium quality (image from description).
Where are generated images and audio stored?
In your files; the chat shows links to them. Files you attach are kept for only one day.
Do the utilities work in the demo?
Yes, within a daily limit: up to €3 a day for AI, parsing and media generation combined. When the limit is used up, you see “The daily demo limit is used up…”. You can continue tomorrow or on a paid plan.
Where to find it and who has access
- Where
- AI consultant — the round button with a robot in the bottom right corner of the desktop
- Other entry points
- “AI tools” in the article and presentation editor, “AI actions” when replying to an email
- Access right
- AI utilities are part of the basic permissions of every role
- Prepare
- an active account (in the demo — a daily limit of up to €3); an image up to 10 MB or audio up to 25 MB
Related topics
- If you need to give the consultant a different task, not one with a file — AI consultant: what to write.
- If you need your own operation in the “AI tools” menu — AI tools in the editor.
- If the utilities should work without you, in an AI employee — AI employees.
