Shot Workspace

The shot workspace is where you spend most of your time. It opens when you click any work item from the home views. Everything related to that one work item lives here: its generations, its takes, its lineage, and the buttons to start more runs.

The top bar of the workspace shows the work item's title, for example "S003 / Shot 001 — Quantum Lab", with its description underneath, and a row of buttons. From left to right:

  • Edit. Open the edit form for the work item.
  • Delete. Only visible if the item has no runs and no artifacts. See Work items for the rule.
  • A status badge showing the item's column on the pipeline board.
  • Node graph and Candidates tabs.
  • + New run. Opens a clean-slate New Run form for this work item.
  • Import. Links existing images or videos from your disk to this work item. See Work items.

By default the workspace opens on the Candidates tab.

The shot workspace for S003 / Shot 001 — Quantum Lab, with the title, Edit button, Review status badge, Node graph and Candidates tabs, + New run and Import buttons, and four takes.

The Candidates tab is the main review surface. Every artifact for this work item is shown as a card, grouped by lineage. A lineage is one creative direction: a particular combination of prompt, inputs, model, and mode. Variations of the same direction sit in the same lineage group. Switch any of those four things and you start a new lineage.

Each lineage group has:

  • A header with the model name and mode (for example "GPT-Image-2 · t2i" for text to image, or "Seedance 2.0 · i2v" for image to video).
  • One artifact card per take, including any take that is still generating.
  • A small + New run card at the end. Clicking it pre-fills the New Run form with this lineage's prompt and inputs, ready to generate another variation in the same group.

Cards wrap onto multiple rows. There is no horizontal carousel and no hidden cards. Everything is visible at once. The vertical scroll bar is the only way to navigate.

A slider at the top of the Candidates tab adjusts card size. Drag it left or right to fit more or fewer cards across the row. Your preference is saved per app, not per project.

The Candidates tab of S002 / Shot 001 — Lecture Hall with two lineage groups, GPT-Image-2 · t2i and Seedance 2.0 · i2v, each ending with a + New run card.

Verdict glyphs on cards

Each card has a small verdict indicator on it:

  • Pending (no verdict yet). The default state right after generation.
  • Accepted. You marked this take as good.
  • Rejected. You marked this take as not usable.
  • Exported. An accepted take that you also exported. Its file has been moved out of the Temp folder.

Click any card to open the detail panel on the right. The detail panel is where you actually change verdicts and run actions. See Reviewing candidates for the full flow.

Generating cards

While a run is in flight, its takes appear as placeholder cards with a spinner. They count down as the model works, then turn into real artifact cards the moment the run completes. If a run fails during generation, no cards are produced. The failure is shown on the node graph instead, with the error message. If Runway refuses a run as soon as you launch it, the New Run form shows the message and no run is recorded.

The Node graph is a directed graph showing how every artifact for this work item was made. It is essentially a visual history. You will see:

  • Input nodes for any reference image or video that was used.
  • Run nodes for every generation, showing the first three lines of the prompt.
  • Artifact nodes for every take that was produced.

Lines connect inputs to runs and runs to the takes they produced. Failed runs show with a red dotted border and the error message visible inside the node. Empty runs (no output artifacts but not failed) show with a plain dotted border so you have a record of the experiment.

The node graph of S003 / Shot 001 — Quantum Lab: one input node on the left, four GPT-Image-2 run nodes in the middle, and their four takes on the right, with the zoom controls and Auto-fit switch at the bottom.

Pan, zoom, and auto-fit

Drag the background to pan. Turn the mouse wheel up or down, or use the + and − buttons, to zoom. Sideways scrolling, such as a sideways swipe on a touchpad, does nothing, and Ctrl plus the mouse wheel does not zoom the app.

The Auto-fit switch, on by default, keeps everything centered and scaled to fit the view as you add more runs. It turns itself off when you drag the background, zoom with the wheel, or use the zoom buttons, so the graph stays where you put it. A simple click never turns it off. Turn it back on to fit everything again.

Clicking nodes

Click an artifact node to open the artifact detail panel, the same one used on the Candidates tab. Click a run node to see its prompt and settings. Clicking only moves the highlight; the graph does not redraw or move. Click anywhere on the empty background to close the panel. (Background drag does not close it, so panning never dismisses your work.)

This is the same panel whether you open it from Candidates or from the Node graph. It is laid out top to bottom as:

  1. Header. An ARTIFACT badge, a close button, the take number as a title, and a subtitle that reads "Model · Mode · Date" (for example "GPT-Image-2 · Text to image · Jun 4, 7:22 PM"). A row of tags on the right shows the provider, aspect ratio, and resolution or duration.
  2. Preview. The image or video itself. Click an image to open it in a full-screen lightbox. Videos play inline with native controls, and the player's own maximize button gives you full screen.
  3. Verdict bar. Three buttons for Accept, Pending, and Reject.
  4. Inputs. Only shown if the run used input references. Each row is clickable and opens the source artifact in the lightbox so you can see what was fed in.
  5. Prompt. The exact prompt text used for this run.
  6. Notes. A free-form text field for your own notes about this take. Saves automatically when you click away or close the panel.
  7. File path. The full path to the file on disk, for when you need to find it in your file explorer.
  8. Action bar. Regenerate, Export, and Delete. Plus an Add to compare set button on the Candidates tab.
The artifact detail panel for Take 001, showing the header with tags, the image preview, the verdict bar with Accept selected, and the input it was made from. The rest of the panel is further down.

The Candidates tab also has a compare set dock that lets you collect several artifacts and view them side by side. Click Add to compare set on the panel to add the current artifact, then open the compare dock at the bottom of the screen.

Imported artifacts

Artifacts that you imported from outside the project look slightly different in the panel: the Regenerate, Export, Delete, and Compare buttons are hidden. Imported files are linked where they are on your disk and count as accepted, so these actions do not apply. 1stUnit never deletes or moves an imported file. You can still change the verdict and write notes.

Drag the left edge of the detail panel to resize it. Your width preference is saved between sessions.

If another program adds takes or changes this work item while the workspace is open, for example an AI agent writing to 1stUnit's database, the Candidates tab and the node graph update on their own. A video that is playing or a panel you have open is not interrupted when other work items change.