Skip to content
Docs0.9GitHubInstall

Using itMarks on photos, video, audio

Marks on photos, video, and audio

Point at a place in an agent's photo, a frame or a stretch of its video, a cue of its narration, or a moment of its audio, with a note the agent gets.

How to tell an agent exactly where something should change in a photo, video, or audio file it sent, for anyone reviewing an agent’s media in the web client or the iPhone app. You markMark. A note you attach to a place: a box on a photo, a frame or a segment of a video, a moment in an audio file, a line of a change. places, write or speak a note on each, and send them as one reply. For a worked example, see Review an agent’s video.

Marks on code work differently: see Changes.

Open the editor from the agent’s reply:

Where
A photo or video Mark in the media viewer. In the web client, also Mark in the message’s menu.
An audio file Mark, the highlighter button on the file

The editor opens on every photo, video, and audio file of that reply, so one send can carry marks on all of them. With more than one, a strip of thumbnails switches between them. The button reads Marks (n) while a draft waits.

Then:

  • Add a mark (below, by kind of media).
  • Select it and write its note, or hold the microphone to speak it.
  • Press Send, add a message if you want one, and press Send marks.

Marks are numbered across the whole reply, in the order shown.

Tool Key in the web client Does
Select V Move and resize a mark
Box B Draw a rectangle
Highlight H Draw a freehand stroke

Two fingers pan and pinch to zoom; in the web client, Ctrl and the wheel zoom.

Marks go on an exact, paused frame. Scrub the filmstrip, or step with Previous frame and Next frame (held, they repeat).

Button Marks
Frame The frame shown
In, then Out A segment from one time to another. Cancel segment abandons one you started.

Draw a Box or Highlight on the paused frame to point at a place in it. A region you draw joins the selected frame or segment mark when that mark has none yet; otherwise it becomes a new frame mark. A selected segment shows In Here and Out Here to move its ends to the current time.

In the viewer, outside the editor, a video plays at 0.25×, 0.5×, 1×, 1.5×, or 2×, and existing marks show as ticks on its filmstrip.

Key in the web editor Does
Space Play or pause
← / → Step a frame
Delete or Backspace Remove the selected mark
Esc Close, keeping the draft

When an agent sends a video with text tracksTrack. A timed text file an agent attaches to a video: its narration, its on-screen text, or its subtitles., the editor lists the track’s cuesCue. One timed line of a track. You can suggest new wording for a cue as a mark. beside the video: each cue’s number, a frame from its start, its times, and its words. The cue playing now is highlighted, and tapping a row goes to it. With several tracks, Track switches between them.

Each cue has a menu:

Item Suggests
Change Your exact wording in place of the cue’s. A hint compares how long your words take to say with the cue’s length.
Remove Dropping the cue
Add after A new cue after this one, with your words
Add at play head A new cue at the current time
Note A note on the cue, with no new wording

A changed cue shows its old text struck through with yours beneath; a removed one is struck through. Each cue takes one mark, besides new cues.

A track is one of three kinds, and the agent is told what a change means for each:

Kind A change means
Narration The voice-over’s words: the agent re-voices the lines and re-times the cut
On-screen text Text drawn on the frames: the agent changes the text shown
Subtitles Captions: the agent changes the caption text

The agent links .srt or .vtt files in the same reply as the video, named after it: intro.narration.srt, intro.onscreen.srt. The kind comes from the name; a file with no such word is subtitles. The agents’ vendetta skill tells them this.

An audio file shows as a waveform. Time marks the current moment; In and Out mark a segment. There is nothing to draw on, and each note names the file it is about, so the agent knows which sample you mean.

  • Select a mark to write its note, up to 1,000 characters. A note may be empty.
  • To speak one, hold the microphone button, or tap it to start and stop. The recording is transcribed by your transcription service when you send, and the agent gets the transcript, marked as spoken. With both a typed and a spoken note, the agent gets both.
  • Spoken notes need transcription set up. Without it the microphone is off and says so.
  • The reply’s own message can be typed or spoken too.

Send shows a summary, such as “3 marks on 2 pictures · 1 spoken”, a box for an optional message, and Send marks.

Your draft, with its pictures and recordings, stays on the device until the agent’s side accepts the send. A send that fails or is refused goes again whole with Retry or Send again; nothing is sent again by itself. If anything in a send is invalid, the whole send fails and nothing reaches the agent.

Your message, then a block listing each marked item and each numbered mark, then one picture per marked photo or frame showing where the marks are, with their numbers and not their notes.

Fix these.

Vendetta marks: the owner marked 2 things you sent. Act on each numbered note. Images 1-2 below show where each mark is, with its number, not its note.

Video 1 intro.mp4
- Sent from: /home/you/site/out/intro.mp4
- Pinned copy: /home/you/.local/state/vendetta/notifications/media/3f2a….mp4 (what the owner saw)
- 1920x1080, 0:12.480 (12480 ms)
- Mark 1 at 0:03.200 (3200 ms); rectangle x 0.1000-0.4200, y 0.2000-0.5500 (pixels 192,216 to 806,594); image 1. Note: "Logo pops in too early"
- Mark 2 from 0:05.000 to 0:07.500 (5000-7500 ms). Note: "Make this part slower"

Photo 2 plot.png
- Sent from: /home/you/site/out/plot.png
- Pinned copy: /home/you/.local/state/vendetta/notifications/media/9b1e….png (what the owner saw)
- 1600x900
- Mark 3 highlight over x 0.2900-0.4900, y 0.0822-0.2378 (pixels 464,74 to 784,214); image 2. Note: "Label overlaps the axis"
  Highlight path (width 0.0200): 0.3000,0.1000 0.3200,0.1200 0.4800,0.2200
  • Each item names the file the agent sent and the pinned copy of it that you saw.
  • Times are exact, in both clock form and milliseconds; regions are given as fractions of the picture and as pixels.
  • An audio item reads “what the owner heard” and gives its length.
  • Cue marks name the track, the cue’s number and times, its current words, and yours:
- Mark 1: change narration cue 3 of "Narration" (intro.narration.srt), from 0:08.200 to 0:10.900 (8200-10900 ms).
  Current:   "Most people who hate them have never even driven one."
  Suggested: "Most of them never drove one."
- Mark 3: add a narration cue to "Narration" (intro.narration.srt) after cue 3, at 0:11.500 (11500 ms).
  Suggested: "Link in bio."

The pictures go to the agent as images, or as private files for a provider that takes none. An agent in a machineMachine. One container that runs a project's agents on a private clone of the project. is told the pinned copy is kept by Vendetta and not readable inside its machine.

Limit
Marks in one send 20
A note 1,000 characters
Marked pictures in one send 10, each up to 2048 pixels long
Suggested cue wording 1,000 characters and 8 lines
Spoken notes and message in one send Together as long as one voice note may be ([plugins.app.voice] max_seconds, 15 minutes by default)
Text tracks SRT or WebVTT, up to 1 MiB and 5,000 cues each; 6 tracks per video
Media that can be marked What the agent sent and Vendetta still keeps. A copy that reads “No longer kept” cannot be marked.

Marks work the same in Simple and Advanced.