Frame.io AI Assistant is a conversational AI layer built directly into Frame.io. AI Assistant understands the context of your projects — your assets, folder structure, comments, metadata, shares, and more — so you can get things one without switching tools or copying anything in.

Organize your project. Describe the folder structure you need and it builds it. Rename assets and apply custom metadata in bulk.

Understand your feedback. Get a summary of reviewer comments across every revision — before you open a single clip. Surface recurring themes and action items.

Optimize your workflow. Improve your process and explore creative directions without leaving Frame.

Frame.io AI Assistant is available to all account types, with usage limits that vary by plan.

Plan	Messages (hourly / daily / monthly)
Free	15 / 30 / 90
Pro, Team, and Business	25 / 100 / 400
Enterprise	100 / 400 / 1,600

Note: All capabilities, limits and behavior of Frame.io AI Assistant may be subject to change during the beta period.

How to activate AI Assistant

To turn on AI Assistant, you must be an Owner or Content Admin of the Frame.io account. Admins can open Account Settings and click on the AI Assistant tab under Account.

Use the toggle to activate AI Assistant. Once toggled on, all members in the account will be able to utilize AI Assistant, but their access to what it can do will be limited to the user's set permission.

Toggle off at any time to disable this feature for the entire account.

Using the AI Assistant

Once the AI Assistant is enabled, you should see a text box at the bottom of any page in your Project space, with the initial prompt asking "What do you need to be done?". Type your first prompt with the AI Assistant to kick off the conversation and open the AI Assistant panel.

You can also find the AI Assistant panel button in the top-right of your Project page or type the Option+5 shortcut to open the panel directly.

Available actions

Here are a handful of example ways to use AI Assistant inside of Frame.io, broken down by categories.

Note: If there is an action not shown in this section below, it still may be a feature to perform in Frame.io. This section will be updated if any additions are made in the future.

1. Finding and understanding your content

Capability	Example prompts
Find assets by location, metadata or search.	"Help me find and organize files" "View recently uploaded files" "Find everything marked Approved"
Look at images, PDFs and videos and describe / summarize them.	"What's in this image?" "Give me the key points of this PDF" "What is this video about?"
Answer questions about assets based on its metadata.	"Who uploaded this?" "What resolution are the selected clips?"
Scout a vague request before acting.	"Is there anything in here we could use for the sizzle reel?"
Find duplicates, the largest files, or files by resolution, type or size.	"Find duplicate uploads in this project" "What are the 20 biggest files?"
Locate files from a list you paste or attach (CSV or text).	"Find these 40 filenames and put them in one folder"
Browse and find folders, including in other projects.	"What's the folder structure here?" "Where's 'Final Deliverables'?"

2. Organizing files and folders

Capability	Example prompts
Create a folder structure.	"Create a folder structure for my project" "Set up Pre-Production / Production / Post-Production"
Sort files by rule (type, date, name pattern, field value).	"Move all PDFs into a Docs folder" "File everything from June into a June folder"
Auto-organize by what's in the files. AI Assistant proposes categories from a sample, sorts everything, creates folders and moves files.	"Help me find and organize files" "Suggest ways to organize my assets" "Split these into interviews, b-roll and graphics"
Move or copy across projects, including whole folder structures.	"Copy this folder to the Nike Spring project"
Rename single files or in bulk: Give rules based on what else is in the Project, including title case, prefixes, find-and-replace, numbering.	"Rename this to Hero_v2_Final" "Rename assets in this folder" "Number these shots 010, 020, 030…" "Rename each clip after what happens in it"
Clean up: delete files matching a rule.	"Delete everything here except the finals"
Undo a delete.	"Restore what I deleted yesterday"
Color-code folders.	"Color all the client folders blue"

3. Metadata and Tagging

Capability	Example prompts
Set up or change fields: create fields and add or rename options.	"Set up or update asset metadata" "Add a 'Shot Type' field with Wide, Medium, Close-up"
Apply the same values to a group.	"Mark everything in this folder Approved" "Set Client to Nike on my selection"
Tag each file from its content (image, video, or document).	"Apply keywords to my assets" "Set Mood for each photo based on the image"
Cull and rate.	"Help me cull and rate my assets"
Classify, then apply a field profile per category.	"Tag product shots as Product/Studio and lifestyle as Lifestyle/Outdoor"
Fill fields from filenames, dates or other fields.	"Set Department from the filename prefix"
Import metadata from a spreadsheet (CSV only).	"Apply the metadata in this CSV"
Fill fields from what's said.	"Set Topic from what each interview is about"
Fill fields from feedback.	"Set Review Status from the open comments"

Note: When AI Assistant fills a field that's hidden in the current view, it turns the field on and shows a card that lets you turn it back off. Fields you hid yourself stay hidden.

4. Commenting and review feedback

Capability	Example prompts
Leave a comment.	"Review and leave notes on assets" "Comment at 0:42 that the logo is cut off and tag Sarah"
Reply to a thread.	"Reply to Mark saying we'll fix it in v3"
Summarize feedback across one or many assets.	"Summarize feedback on my assets" "Unresolved comments in this share"
Catch up on what you haven't read, including mentions of you.	"What feedback haven't I read yet?"
Batch-create comments by rule.	"Add 'needs color pass' to every clip in the Grade folder"
Time coded comments at every mention in speech or what's on screen.	"Leave a comment wherever someone says 'pricing'" "Find when the logo appears and leave a comment there"
Batch-resolve, reopen or delete comments by author, text, status or age.	"Resolve all of Mark's comments older than a week" "Delete my test comments"
Creative review of images, scored 1–10 on composition, color, lighting, brand and more.	"Review this key art against the brief"

Note: Editing a single comment isn't supported.

5. Transcripts and spoken content

Capability	Example prompts
Transcribe video or audio, with speakers identified.	"Transcribe all the interviews" "Add captions to this video"
See which files have, or still need, transcripts.	"Which videos still need transcribing?"
Summarize a recording, pull out action items, attribute speakers.	"Summarize a transcript" "List the decisions from this meeting"
Find where a topic was mentioned across a project or the whole account.	"Which videos mention the rebrand?"
Find every occurrence, with timecodes and counts.	"Every time 'Acme' is said, with timecodes"
Speaker analytics.	"How much did each person talk in this panel?"
Reports from spoken content.	"Make a spreadsheet of every quote about pricing"

Note: If a file has no transcript, AI Assistant offers to start one. If the question can be answered visually, it looks at the frames meanwhile.

6. Video and visual understanding

Capability	Example prompts
Precise questions about timing, motion or audio in a video.	"When does the product first appear?" "When does she start speaking?"
Grab exact stills from a video, or turntable frames from a 3D model.	"Grab a still at 1:23"
Same thumbnail frame for many videos.	"Use 0:05 as the thumbnail for all these"
Best thumbnail per clip.	"Pick the best frame for each clip's thumbnail"
Find a moment, then act on it: set a thumbnail, grab a still, comment.	"Find the hero shot and make it the thumbnail"

Note: Audio-only files can't be listened to; AI Assistant relies on their transcripts.

7. Reports and documents

Capability	Example prompts
Export data about assets as CSV, Markdown or text.	"CSV of every asset with Status and Client"
Summarize documents: Word, Excel, PDF.	"PDF summary of the client feedback"
Interactive reports: dashboards and sortable tables that open live in the Frame.io viewer, where people can comment.	"Interactive dashboard of this project's review progress"
Reformat or merge your data with project data.	"Turn this CSV into a styled client report"
Quick in-chat reports (counts, distributions, audits).	"How many files of each type are in this project?"

Note: AI Assistant only supports CSV, text, or PDF files; Excel or Word files can't be read. Up to 5 files per action, 25 MB each and 50 MB in total.

8. Projects and people

Capability	Example prompts
Create a project.	"Create a project called 'Q4 Campaign'"
Find a project and work across projects.	"Copy these into the Nike Spring project"
Check what you're allowed to do in a project.	"Can I make public shares here?"
Find people and groups for mentions, person fields and reviewers.	"Who's on this project?"

Unavailable actions

Below are the general actions currently unavailable to in the AI Assistant.

Hearing audio: AI Assistant can't listen to audio-only files. A video's soundtrack is only heard through Video analysis.
Excel or Word input: it can't read these as attachments.
Visual grouping: it can't group files into version stacks by how they look.
Exhaustive search: search returns the top 200 matches; full scans need Bulk.
Share contents: shares can't be searched by content, and "who hasn't viewed" can't be answered.
AI attribution: nothing is marked "made by Frame.io".

Skills: saved, repeatable workflows

Skills can be used as identifiable workflows that the Frame.io AI Assistant can come back to whenever it is needed. Below are the list of Skill actions that can be prompted.

Capability	Example prompts
Save a workflow as a Skill and run it later with `/`.	"Save this as a skill called 'Client delivery prep'"
Offer to save repeated requests.	(AI Assistant notices you've asked something similar before)
List, view or edit your Skills.	"Show my skills" "Update my review skill to include watermarks"
Delete a Skill.	"Delete my old review skill"
Self-maintaining Skills. When a saved Skill refers to an outdated tool name, AI Assistant fixes it and re-saves the Skill without asking, then says so.	(automatic action)
Starter Skills: Organize project media, Summarize review feedback, Prepare a client review share.	`/` in the composer

Built-in playbooks

AI Assistant loads these playbooks automatically when a request matches. A small status line appears when it does.

Playbook	Used when you ask to…
Bulk asset actions	organize, clean up, rename, or act on "only the ones that match X"
Bulk metadata updates	tag, rate or classify many files, or apply a CSV
Versioning assets	stack, merge, unstack, or clean up versions
Shares	create, audit, or bulk-change shares
Transcriptions	find what was said, summarize, transcribe, timecode
Creative media review	score images on a 1–10 rubric

FAQ

Q: Which AI model is used, and who owns AI-generated outputs and derivative content?

A: We use the latest Claude models to power the product. AI-generated outputs and derivative content is owned by you (the customer) and we do not train on your content.

Q: How many assets can be selected in a single prompt with Frame.io AI Assistant?

A: Up to 3,000 assets in a Project or 1,000 assets in a Share Link.

Q: Do AI governance controls exist for data retention, deletion, history management, and disabling AI functionality?

A: We allow all admins to disable AI Assistant at the account level for all users.