Runway Gen3 Image-to-Video (fal)

Runway is the New York company that, in a roundabout way, is why local generation exists at all - it published the Stable Diffusion 1.5 weights back in 2022. These days it's a closed commercial video shop, and Gen-3 was its breakout image-to-video model. This node lets you animate a still with Gen-3 straight from ComfyUI, through fal.ai, without touching Runway's own subscription app.

It's a hosted call: no weights locally, no GPU used on your end, billed per generation on fal. Runway's models are closed, so the API is the only door in.

How it works

You give it a starting image and a prompt describing the motion, and Gen-3 generates a short clip that begins from your frame and moves according to the prompt. The still anchors the look; the prompt drives the camera and the action. It's the standard image-to-video setup, and Gen-3 was one of the models that made that workflow feel usable.

The inputs and outputs that matter

Short list here, which is part of the appeal:

• prompt (multiline) - describe the motion, not the image. The picture is already fixed by your input frame, so spend the prompt on what moves and how the camera behaves.
• image (IMAGE) - the first frame. This sets composition, subject, and style. A clean, high-quality still gives you a much better clip.
• duration - 5 or 10 seconds. Ten costs more and gives the motion more room; five is the cheaper iteration length.

Output is a STRING - a URL to the finished video on fal's servers, not a VIDEO you can preview in the node. To watch or keep it, wire the string into the pack's Fal Save Media from URL node (it downloads to your output/ folder) or into a downstream fal node's *_direct_url input to keep chaining. fal result URLs expire, so save anything worth keeping.

How to install it

From gokayfem's ComfyUI-fal-API pack. ComfyUI Manager: search ComfyUI-fal-API, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/gokayfem/ComfyUI-fal-API.git
pip install -r ComfyUI-fal-API/requirements.txt

Add your fal key (copy config.ini.example → config.ini, paste from fal.ai/dashboard/keys, or export FAL_KEY=your_key) and restart. Nothing downloads locally.

Common issues & troubleshooting

You wired the output into a video preview and got nothing. The output is a URL string, not a VIDEO. Route it through Fal Save Media from URL first, then preview the downloaded file.

Motion ignores your prompt. Describe the motion plainly and keep the image description out of the prompt - restating what's already in the frame confuses the steering. If it's still static, a 10-second duration gives the model more to work with.

The clip drifts off your starting image. Image-to-video always trades some fidelity for movement. A sharper, less ambiguous input frame holds better; busy or low-res stills wander.

Insufficient funds / billing lag. fal bills per call and is a legitimate platform, but top-ups have occasionally been slow to post for people. If credits aren't showing, wait for them rather than paying twice. Any fal-side error (quota, policy, validation) comes through as readable text in ComfyUI.