Skip to main content
The Tella Agent is the chat in the sidebar on the right of the editor. Everyone has it, with no beta access to request; sending messages needs a paid plan. Describe the change you want in plain language and the agent edits your video step by step while you watch.
The Tella Agent is in public beta. Send feedback on what works, what doesn’t and what you’d like next to support@tella.com.

What can the Tella Agent do?

The agent uses the same tools you use in the editor. You can ask it things like:
  • “Cut the filler words and the long pauses.”
  • “Remove the section where I talk about pricing.”
  • “Zoom in when I click the Settings button.”
  • “Use a split layout for the demo part and full camera for the intro.”
  • “Add subtitles and translate them to Spanish.”
  • “Write a title and chapters for this video.”
  • “Make a Short from this video.”
It reads your transcript and timeline and shows each change in the editor as it goes. You can keep working while it runs, and ask it to undo a change or try something different.

Open the Tella Agent

  1. Open a video in the editor.
  2. If the sidebar on the right is hidden, click Toggle sidebar at the right end of the playbar above the timeline, or press Shift + S.
  3. Type what you’d like to change in the message box at the bottom of the sidebar and press Enter.
While the agent is working with the sidebar closed, the Toggle sidebar button pulses so you know it’s still busy.

Stop the agent or change course while it works

  • Stop it: while the agent works, the send button turns into a square Stop assistant button. Click it to end the current turn.
  • Send a follow-up: keep typing in the message box (it says Ask for another edit) and press Enter. Your message is marked Waiting until the agent takes it in, and it then steers what the agent does next.

Start a new video from My videos

You can start a new video with the agent: the What do you want to make? box at the top of My videos is the agent’s composer.
  1. Go to My videos, or click your workspace name and choose New video.
  2. Type what you want to make, for example “Make an animated title for my intro”, and press Enter. Tella creates a new video and opens it in the editor, where the agent starts working.
Under the box you can also click Record to record the first clip, Skills to pick a skill, or or start from a blank canvas to open an empty video in the editor. Click + in the box to upload a file. Under Examples, hover a card to preview what it makes, and click it to fill the box with its prompt. Scroll the row sideways to see more. Some cards are made with a community skill, like Make a stop motion intro: clicking one also adds that skill to the box as a pill. Click Browse skills to see every community skill, or click Examples to fold the row away. Click Minimise to collapse the box to a single Ask Tella to create a video bar; click the bar to open it again.

Can I turn off the agent or go back to the old editor?

No. There’s no setting to turn the Tella Agent off and no older editor to switch back to. To edit without the agent, click Show transcript next to the message box and cut words in the transcript instead. See Edit with the transcript. The sidebar remembers your choice in that browser: once you switch to the transcript, it opens on the transcript for your next video too, until you click Show chat to go back to the agent.

Do I need beta access or a setting to turn the agent on?

No. The agent is in the editor for everyone: open a video and the chat is in the sidebar on the right. If the message box shows The Tella Agent is included with paid plans, with monthly credits. and an Upgrade button, your workspace has no paid plan or free trial, and you can’t send the agent messages until you subscribe. See How much does the Tella Agent cost?.

Answer questions and fine-tune with cards

Sometimes the agent puts an interactive card in the chat instead of asking in plain text.

Question cards

When the agent asks you to pick one answer, it shows a card of lettered options (A, B, C…). Click an option, or press its letter on your keyboard, to answer right away. To answer in your own words, click Something else… in the last row and type your reply. When the agent needs a few details before it edits, like how long a pause must be before it’s cut, it shows a card with sliders, choices, switches or a text box. It fills in its own suggestions, so you change only what you want.
  • Some choices show a picture or play a sound so you can compare them before you pick.
  • With three or more questions, the card shows one question at a time. Click Next and Back to move between them; picking a single choice moves on by itself.
  • Click Continue (or the button the agent named) to send your answers, or Skip to skip the questions.
  • Or type a reply in the message box.
Once you answer, the card shrinks to one line per question with your answer.

File upload cards

When the agent needs a file that only you have, like your logo, a screenshot, a voiceover or a brand font, it shows a card that says which file it needs. It waits for your answer before it goes on.
  • Click Choose a file and pick it, or drop the file on the card or anywhere in the chat.
  • Click Skip if you don’t have it. The agent goes on without it or asks you what to do.
  • If the agent asked for certain file types, a file of another type isn’t accepted, and the card lists the types it takes.
If you send a new message instead, the agent stops waiting for the file.

Live controls

For small adjustments, like a sound effect’s volume, caption size, layout padding or zoom strength, the agent can show a card of controls marked Applies live. Moving a control changes your video right away, without sending a message to the agent.
  • Sound effect and caption controls work like the editor’s own: you can undo them as usual. They only work while the video is open in the editor.
  • Other controls save when you let go and show Saved. These changes aren’t in the editor’s undo history, so click Reset to go back to the value the card started with.

AI video clip cards

Before the agent generates an AI video clip, it shows a card with the model, the clip length, the resolution and about how many credits the clip will use, followed by the prompt it will use.
  • Click Generate clip to start the clip.
  • Click Cancel to skip it. The agent can then offer a shorter clip, a cheaper model or no generated video.
If the agent wants several clips at once, you get one card per clip. To pick the model yourself, name it in your message.

Upgrade and credit cards

If you ask the agent to upgrade or buy credits, or something you asked for needs a paid plan or more credits, the agent can show a checkout card in the chat. It doesn’t suggest upgrades unless you ask or you’ve hit a limit.
  • Upgrade to … cards show the plan’s price. Click Upgrade to pay in Stripe Checkout, or Compare plans to open Plans & Billing. If your workspace already pays, you change plans on Plans & Billing instead.
  • Buy agent credits cards show how many credits are left and a tile for each one-time top-up. Pick a tile, then click Buy … credits to pay in Stripe Checkout.
Stripe opens in a new tab, and the agent never sees your payment details. Only workspace owners can pay. Other members see Only a workspace owner can buy credits. Ask one to add more.

Point the agent at part of your transcript

To be specific about which part of the video you mean, select words in the transcript and click Add to chat. The selection appears above the message box as Transcript with its word count, and the agent edits that part.
While the sidebar shows your transcript, the message box under it shrinks to one line with only + and the send button. Click it to open the full message box with the model menu and credits button. It shrinks again when you click away, unless you’ve typed a message, attached something or added transcript text.

Attach files, videos and playlists

Click Add (+) in the message box to give the agent more to work with:
  • Attach…: add documents (like a script or brief), images, videos or audio for the agent to read and use.
  • Folder…: connect a folder from your computer so the agent can look through it and use only the files it needs. See Connect a folder from your computer.
  • Video or playlist…: reference another of your videos, or a playlist in your workspace, for example to reuse a clip or match its style. You can also type @ in the message box to search for one. Only videos you created or own appear in the search.
Videos and images you attach are also added to this video’s media library, so you can place them yourself too. Open Overlay → Media → Browse all and choose Video. Documents aren’t added to the library. You can also ask the agent to save media it uploads or generates to this library. To hand the agent something already in your media library, click Media in the editor’s left toolbar, then drag a tile onto the chat, or open the tile’s ⋯ menu and choose Add to chat. Add to chat opens the chat if it’s closed. The item appears in the message box with its thumbnail and name, and the agent can place it straight away. Drops only work while the sidebar shows the chat, not your transcript. Media that’s still generating or failed to generate can’t be attached.
After you send a message, click an image you attached to open it full screen. Use the arrows to page through it along with the media the agent has shown in the chat. Images in older messages can’t be opened.

Connect a folder from your computer

Click Add (+) → Folder… in the message box and pick a folder, like a folder of screen recordings, B-roll or brand assets. The agent can then:
  • List the files in the folder.
  • Preview videos, images and audio without uploading them, for example a contact sheet of frames from a video, so it can pick the right file.
  • Upload only the files it reads for your edit. Other files stay on your computer.
In Chrome and other Chromium browsers, the folder stays connected after you reload the page. The next time the agent opens it, it asks Open your folder “…”?: click Allow to let it in, or Don’t allow. In other browsers, the folder stays connected until you leave the page; if the agent asks for it later, connect it again from + → Folder….

Get documents from the agent

Ask the agent for a file, like “make me a PDF of the script”, “export the chapters as a CSV”, “give me the captions as an SRT file” or “turn this into a slide deck”, and it shares it in the chat as a card with the file’s type, name and size.
  • Click the file name to open a PDF, CSV, TXT, SRT, VTT or Markdown file in a new tab. Other types download. SRT and VTT caption files show as TXT on the card.
  • Click the download icon to download any file.
The agent can share PDF, CSV, TXT, SRT and VTT captions, Markdown, ZIP, DOCX, PPTX and XLSX files, up to 50 MB each. The files stay in the chat, so you can come back and download them later.

Record a clip from the chat

Ask the agent to write you a script, or to help you record something new, and it can show a recorder card in the chat: Record a clip, or Record with this script when it wrote one.
  • Click Show full script to read the whole script, or Copy to copy it.
  • Click Go to recorder to open the recorder for this video, the same as the editor’s own record button. If the card has a script, it’s put in your speaker notes and the notes open, ready to read while you record.
The script replaces whatever was in your speaker notes before. The Mac and Windows apps get the script too once they’re up to date; older versions open the recorder without it, so use Copy and paste it into your notes. Go to recorder only appears in a video’s own chat. Elsewhere the card shows the script without the button.

Choose a model

Click the model button in the message box (it shows the model and level, like Sonnet 5.5 Medium) to pick the AI model the agent uses and how much it thinks before answering. Your choice applies from your next message. You can’t change it while the agent is working. Type in Search models… at the top of the menu to filter the list. The slider at the bottom of the menu sets how hard the model thinks. It’s called Thinking level for Claude models and Reasoning effort for OpenAI models, and goes Low, Medium, High, Extra high, Max. It starts at Medium. GPT-6.1 Terra goes up to High. Larger models and higher levels handle harder edits but use more credits per message. To run the Claude or OpenAI models on your own ChatGPT or Claude plan, see Use your ChatGPT or Claude plan with the Tella Agent. Claude Fable 5.2 is the most capable model for long, multi-step edits, and it uses more credits per message than the other models, so the menu marks it High credit usage. The label goes away once you’ve connected a Claude plan, because your plan pays for the model. Anthropic only offers Claude Fable 5.2 to workspaces that allow data retention, so it isn’t in the model menu if your workspace is set up for zero data retention. Pick another Claude model instead. Claude Fable 5.1 is no longer in the model menu. If it was your last pick, the agent switches back to the default, Claude Sonnet 5.5.

Memory: what the agent remembers

The agent remembers preferences you tell it, like “always use my brand color for text” or “I prefer jump cuts over transitions”, so you don’t have to repeat them. Click Add (+) → Memory to see what it has saved:
  • About you: preferences that only apply to you.
  • This workspace: preferences shared with everyone in the workspace.
Remove anything you no longer want it to remember from the same window.

Skills: reusable editing workflows

A skill is a saved set of instructions for an editing job you do often, like cleaning up a demo or preparing a social clip. Run one and the agent follows the same steps every time, without a long prompt.
To run a skill, type / in the message box and pick it from the list, or open the skill library and click Use in chat. Open the skill library with the Skills button at the top of the sidebar, or from Add (+) → Skills. Skills are grouped by who can use them:
  • Personal: skills only you can use.
  • Workspace: skills shared with everyone in the workspace, alongside Tella’s built-in skills.
Drag a skill between the two groups to share it or keep it to yourself. A personal skill takes priority over a workspace skill with the same name. To share a skill outside your workspace, click Share with the community to send it to Tella for review and list it on the community skills page: see Share a skill or publish it to the community. In a new chat, the pills under the agent’s welcome message run a built-in skill in one click: Plan a Video, Auto Cut, Auto Layouts and Remove Mistakes. Clicking one sends the request right away.

Plan a video before you record

Click Plan a Video, or run the built-in tella-plan-video skill, when you want help with what to say and show before you record.
  1. The agent asks what the video is for.
  2. It writes the one-line point, an outline, a word-for-word script sized to the length you want, and a shot-by-shot storyboard.
  3. It waits for your go before it changes anything, so you can ask for changes first.
It also works on a long, rambling take: ask for a tighter version to record again. To read the script while you record, ask the agent for a recorder card (see Record a clip from the chat). For a launch, promo or feature announcement, run tella-launch-video. For a “what’s new” video from your release notes, run tella-changelog-video.

Make a Short, Reel or TikTok

Ask the agent to “make a Short from this video”, or run the built-in tella-short skill, to turn a video into a vertical clip for Shorts, Reels or TikTok.
  1. The agent picks the strongest 15 to 60 seconds and shows you a scene-by-scene plan: the segment, the opening hook and each shot.
  2. Reply to approve the plan or ask for changes. Nothing is edited until you do.
  3. The agent makes a trimmed copy of your video, tightens it, switches it to 1080×1920, turns captions on and lays it out at Shorts pace, then checks every cut.
Your original video is never changed. To steer the result, tell the agent the time range, chapter, length or platform you want.

Save an edit as a skill

After the agent finishes a multi-step edit, it may offer to turn it into a skill: click Save as skill, or Don’t suggest to stop suggestions in that chat. You can also ask it directly, for example “Save what you just did as a skill.” The agent asks for your approval before it saves or changes a skill. For writing, editing and deleting your own skills, see Create a skill for the Tella Agent. A skill can include reference files, like a video or image that shows the style you want. Attach the file or point the agent to a clip and ask it to add it to the skill. Each file can be up to 10 MB, so the agent trims or re-encodes longer videos first.

Add a skill from GitHub

  1. In the skill library, click Add a skill from GitHub.
  2. Paste the repository link (for example https://github.com/owner/repo).
  3. Review what the skill contains, choose Personal or Workspace, and click Trust & install.
Only install skills from sources you trust: a skill can run code in the agent’s sandbox. You need to be a workspace owner or member to add skills or change workspace skills.

Use a community skill in one chat

Community skills are skills shared by the Tella community, like animated transitions and title styles. Browse them at tella.com/skills. You can try one in a chat without adding it to your library.
  1. Open a skill on tella.com/skills and click Copy under Use it in Tella to copy its ID, for example tellahq/community-skills/kaleidoscope.
  2. Paste it into the agent’s message box. The skill appears as a pill in the message box.
  3. Write what you want, for example “Use this for the transitions between my chapters”, and send.
The skill only applies to that chat, and the message shows a card with its name, author and description. To keep it, click Save to my skills on the card: it’s added to your Personal skills and the card says In your skills. If the pasted text doesn’t match a community skill, it stays in the message box as plain text. When you ask for a motion graphic, transition, title card or animated diagram that none of your skills covers, the agent may suggest a community skill that fits, with a link to it. Paste its ID into the chat to try it, or ask the agent to build the effect itself.

Rate a reply

Hover over a finished reply from the agent to see the thumbs-up (Good response) and thumbs-down (Bad response) buttons next to Copy response. Click one to tell us whether the reply helped. The thumb you chose stays filled in, even after you reload the page or come back to the chat later. To change your rating, click the other thumb. Only you see your ratings.

Chats

Each video keeps its own chat history. Start a New chat for a fresh conversation, or open the chat list to go back to, rename or delete an earlier one. Everyone editing a video shares its chat with the agent, so co-editors see the same conversation.

How much does the Tella Agent cost?

The Tella Agent is included with paid plans (Pro, Premium and Enterprise) and free trials, and runs on Credits. Each seat on Pro includes 9,000 credits a month and each seat on Premium includes 18,000 credits, pooled across the workspace. Included credits reset every month and don’t roll over, and workspace owners can top up anytime. Each message uses credits based on how much work the agent does and the model it uses. A free trial includes 9,000 credits, so you can try the agent before you subscribe. They expire when the 7-day trial ends.

The agent says I’m out of credits

When your workspace runs out of credits, the message box turns yellow and says You’re out of credits. Add more to keep using the agent. A workspace owner can buy more from the Add credits button next to the message box or in Settings → Plans & Billing. See Credits.
Last modified on November 21, 2026