Your First Workflow

Overview

Here's what a typical video generation workflow looks like:

1. Generate an image from a prompt

2. Turn that image into a video clip

3. Extract the last frame of the clip

4. Generate a second (and third) video clip from that frame

5. Assemble all clips in the Editor

6. Add captions

By the end, you'll have a polished, seamlessly stitched video — generated entirely on autopilot.

---

Step 1: Create a New Workflow

Go to the Workflows tab in the left sidebar. Click New Workflow and give it a name.

The Workflows tab showing the New Workflow button and folder organization.

---

Step 2: Understanding Nodes

Nodes are the building blocks of your workflow. To add a node, click the + button (or right-click on the canvas) to open the node menu.

Nodes are organized into categories:

The node menu showing all available node categories.

For most video generation workflows, you'll primarily use fal.ai and Editor nodes.

---

Step 3: Add Your First Node — Generate an Image

Open the node menu → fal.ai → Create Image. Click to place it on the canvas.

Each node has a panel on the right where you configure its settings. For Create Image, the key settings are:

• Prompt — describe the image you want to generate

• Aspect Ratio — set to 9:16 for vertical video

• Resolution — choose your output quality

• Model — pick from Banana, Dream, or GPT image models

The Create Image node with prompt and resolution settings filled in.

---

Step 4: Add an Image-to-Video Node

Next, you want to turn that generated image into a video clip.

Open the node menu → fal.ai → Image to Video. Place it on the canvas to the right of your Create Image node.

Map the nodes together

This is how you connect one node's output to another node's input — called mapping.

1. In the Image to Video node settings, find the Start Image field

2. Click the field — you'll see a list of available outputs from other nodes

3. Select Create Image → images

A line will appear connecting the two nodes on the canvas.

Connecting the image output of Create Image into the Start Image input of Image to Video.

Now configure the Image to Video node:

• Prompt — describe the motion and action in the clip. For lip-sync videos, describe the character's expressions, emotions, and dialogue in detail.

• Duration — keep clips at 10 seconds or under for best lip-sync quality. For longer videos, split them into multiple 10-second segments.

• Generate Audio — toggle on if you want the model to generate audio alongside the video.

Image to Video settings with prompt and duration configured.

---

Step 5: Extract the Last Frame

To create a seamless continuation, you need the last frame of your first clip as the starting image for the next one.

Open the node menu → fal.ai → Extract Frame. Connect it after your Image to Video node:

• Input → the video output from Image to Video

• Frame → set to Last

The output of this node is an image — the final frame of your clip — ready to feed into the next Image to Video node.

Extract Frame connected between two Image to Video nodes for seamless continuation.

Repeat steps 4–5 to generate your second and third clips, each starting from the last frame of the previous one.

---

Step 6: Assemble Clips in the Editor

Once you have all your video clips, it's time to stitch them together.

Open the node menu → Editor. The Editor node works differently from other nodes — it has no inputs by default. You define its inputs from inside the Editor itself using Placeholders.

Add Placeholders

1. Click Open Editor on the Editor node to enter the timeline view

2. Click Add Media → Placeholder

3. Add one placeholder per video clip (e.g. three placeholders for three clips)

4. Arrange them in order on the timeline

5. Set the Duration for each placeholder to match the expected clip length (e.g. 10 seconds)

Three video placeholders arranged in sequence on the Editor timeline.

Once you save and return to the canvas, the Editor node will now show input handles for each placeholder you created.

Map your clips

Connect each video output (from your Image to Video nodes) to the corresponding placeholder input on the Editor node.

Each video clip mapped to its placeholder input on the Editor node.

Add static layers

You can also add static elements directly inside the Editor (no placeholder needed):

• Logo or watermark — upload an image and position it on top

• End card / body — add a static image or video at the end of the timeline

• Use the layer order to control what appears on top: higher layers overlay lower ones

A logo image layer sitting above the video clips in the Editor timeline.

---

Step 7: Add Captions (Optional)

If you want subtitles, add a Captions node after the Editor.

Open the node menu → Captions. Connect the Editor's video output to the Captions input, then configure:

• Language — select the correct language for accurate transcription

• Position — set how high the subtitles appear relative to the safe zone

• Template — choose from multiple subtitle style presets

Captions node with language, position, and template configured.

---

Step 8: Run Your Workflow

Test individual nodes

While building, you can test any single node without running the whole workflow. Click the Run button on a specific node — it will execute just that step and show you the output immediately.

Use this to verify your prompts and settings before committing to a full run.

The Run button on an individual node for quick testing.

Run the full workflow

Once everything is connected and configured, click the Run button at the top of the canvas to execute all nodes in sequence.

You'll see a new entry appear in Run History. Click into it to view each node's output, preview the final video, or download it.

The Run History tab showing a completed workflow run with downloadable output.

---

What's Next?

Now that you have a working workflow, explore:

• Collections — run your workflow multiple times with different inputs automatically

• Run History — re-run failed nodes, revert to previous versions

• Editor — dive deeper into timeline layers and dynamic durations