How do I create my first Martini project?
Short answer
Click New Project on the dashboard, name it, and you land on an empty canvas. Drag your first node from the left toolbar — an Image node is the simplest start — pick a model, write a prompt, and click Run. The output appears on the node within seconds. From there, drag more nodes and connect them to build a workflow.
Create the project
Sign in and you land on the dashboard. The dashboard lists your projects, recently opened canvases, and a New Project button in the top right (or center on a fresh account). Click New Project. A modal asks for a project name and an optional description. Pick a name that describes the brief — Client X Launch Video, Product Photoshoot Variants, Music Video R1 — and click Create. The new project opens to an empty canvas.
Workspace members see a workspace selector before naming the project — pick the personal space for solo work or a workspace for shared work. Once created, the project lives in the chosen context and inherits its sharing defaults. Personal projects stay private; workspace projects default to workspace-internal visibility.
Drag your first node
The empty canvas has a left toolbar with node categories — Image, Video, Audio, Text, Script, Tool. Click and drag a node onto the canvas. Image is the easiest first node: drop it anywhere, and a panel appears showing the model picker, prompt field, and parameters (aspect ratio, resolution, batch size). Pick a model, write a clear 1-2 sentence prompt, and click Run.
The first run takes a few seconds. The output preview appears inside the node when ready. If the output is not what you wanted, edit the prompt and click Run again — Martini does not lock you into a single output. Each run is a fresh generation and costs credits, so iterate at lower resolution or fewer batch outputs first if you are budget-conscious.
Connect a second node
Drag a second node onto the canvas — a Video node, for example. The Video node has an input port on the left side where it accepts a reference image. Drag a connection from the output of the first Image node to that input port. The connection turns green when valid. Pick a video model, add a video prompt (for example, the camera dollies forward slowly), and click Run.
The video model uses the upstream image as the starting frame and the prompt as the motion description. This is image-to-video — the most common two-node workflow on Martini. Once the video returns, you can chain a third node: a lip-sync, an upscale tool, an audio mix. Each new node adds a step to the pipeline.
Save, name, and share
Martini auto-saves the canvas as you work — there is no explicit Save button. The project name and description are editable from the top bar at any time. The Share button next to the project name controls visibility — link-only, workspace-wide, or private. The right-side panel shows generation history, credit spend, and exports for the current project.
From here, the project becomes your iteration surface. Add nodes, run generations, branch off variants, and refine. The canvas is infinite, so a single project can hold a full client deliverable with dozens of nodes. Use the Project menu to export, duplicate, or archive.
Examples
- New project for a product launch — start with one Image node, then chain a Video node for an ad clip.
- Music video draft — Image node for the cover, Audio node for the track, Video node for the hero shot.
- Client review project — share via link-only so the client can preview without signing up.
- Workspace project for a team brief — created in the workspace context so all members see it.
Edge cases
- A project created in the personal space cannot be moved to a workspace silently — duplicate and re-create instead.
- The Free plan limits how many concurrent projects you can keep active; archive old ones to free slots.
- Disconnected nodes can still be run but produce isolated outputs — the canvas is non-linear by design.
- Auto-save is server-side; closing the browser tab without losing changes is fine.
What to do next
- Click New Project from the dashboard and name your project.
- Drag an Image node onto the canvas, pick a model, write a prompt, and click Run.
- Connect a Video node to chain image-to-video for your first multi-step workflow.
- Read the canvas-basics article for a deeper tour of the canvas interface.
Related help articles
Still need help? Contact support.