AI Meeting Summary from Video

Turn any recorded video call into a short, named summary.

Upload a Zoom, Teams, or Meet recording. The workflow reads participant names straight off the on-screen name tags (the labels under each video tile), transcribes the audio with Whisper, and merges both into a casual recap. You get three things back: a quick gist of the call, a who-said-what breakdown grouped by each person's real name, and an action-item checklist with owners and deadlines where mentioned.

Names come from what the camera sees, not from voice recognition or guesswork. If someone's name tag is visible, it shows up in the summary.

How do you summarize a video call with participant names using AI?

Upload your meeting recording and hit run. Gemini Flash vision reads the on-screen name tags directly from the video frames while Whisper transcribes the audio. An LLM then combines both into a recap grouped by each participant's real name, written in a casual tone you can skim in under a minute.

Video file
Drop in your call recording. Any standard video format works: Zoom cloud exports, Teams recordings, Google Meet files, or a screen recording. One file goes in, and the workflow splits it into video frames (for reading names) and an audio track (for the transcript).

Name detection
Gemini Flash vision scans sampled frames from the video and reads the text labels that appear under each person's video tile. It only reads what it can see on screen. If a participant's camera was off or their name tag was cropped out, they won't appear by name. The workflow falls back to "Speaker 1, Speaker 2" labels when no tags are visible.

Transcription
Whisper (large model) handles the audio-to-text step. Language detection is automatic. You don't need to set anything here.

Output format
The summary is written like a quick Slack message to a teammate, not a formal meeting-minutes document. It has three sections: the gist (3-4 sentences on what the call covered), who said what (bullet points grouped by name), and action items (checklist format with owners and deadlines where mentioned).

What is an AI meeting summary with named participants good for?

Best for catching up on a recorded call without watching the whole thing. The named breakdown means you can jump to what a specific person said or committed to, instead of scanning a wall of undifferentiated transcript text.

Missed a standup? A client call that ran long? A brainstorm you were double-booked against? Upload the recording and get the key points in a format you can read in 60 seconds.

The real value is the name attribution. Generic transcription tools give you a block of text with no speaker labels, or they guess speakers from voice patterns (and get it wrong). This workflow reads the actual name tags from the video, so attribution is based on what was displayed on screen during the call.

Where it works best: any video call platform that shows participant name tags on screen. Zoom, Microsoft Teams, Google Meet, Webex. If the recording captured the gallery view with visible names, you're set.

Where it won't help: audio-only recordings (no video frames to read names from), calls where cameras were off the entire time, or platforms that don't display name labels. In those cases the summary still generates, but participants get generic "Speaker 1" labels instead of real names.

FAQ

Does this work with Zoom, Teams, and Google Meet recordings?
Yes. Any recording that captured the video call interface with visible name tags works. The model reads the on-screen text labels, so it doesn't matter which platform produced the file. If you can see name tags when you play the video, the workflow can read them too.

What happens if some participants had their cameras off?
The workflow reads names from on-screen labels. If a participant's camera was off and their name tag wasn't visible in the recorded frames, they won't be identified by name. Their contributions still appear in the summary under a generic "Speaker" label based on transcript context.

How long of a meeting can this handle?
The main constraint is Whisper's processing time on the audio track. Meetings up to an hour process well. For longer recordings, processing time increases but the workflow still runs. The summary output stays concise regardless of input length.

Can I change the summary tone from casual to formal?
The default output is written in a casual, colleague-to-colleague style. To change the tone, expand the packaged node and edit the LLM system prompt inside. For most teams, the casual format is faster to scan than formal minutes.

How do I run an AI meeting summarizer online?
You can run this meeting summarizer online through Floyo. No installation, no setup. Open the workflow in your browser, upload your recording, and hit run. Free to try.