MiniMax Hailuo 2.3 Fast

Hailuo 2.3 Fast is the i2v-first lane of MiniMax's video family, and the node layout tells you immediately: frameImages is the required input. Unlike the standard Hailuo 2.3 node, where you prompt from scratch, this one assumes you already have an image and want it moving - fast. It's the node for the workflow where you nail a still with an image model, then hand it to Hailuo to bring to life.

The "Fast" in the name is doing real work, too. Same Hailuo 2.3 generation underneath, but tuned for quicker turnaround - fewer total steps in the cloud means lower cost and shorter queue, at the cost of some polish on complex motion. For iterating on a shot, it's the right tool. When you've found the version you love and want the best render, you'd move up to the full 2.3 node with the same frame and prompt.

What you set

• frameImages (required) - your starting still, as IMAGE. This is the star of the show.
• positivePrompt (optional) - what should happen. "The character turns and waves, camera slowly pushes in" is the kind of direction this node wants.
• width / height - keep them matching your input frame's aspect.
• duration toggle + duration_value - seconds at the model's fps; off uses the default.
• providerSettings.minimax.promptOptimizer - prompt rewriting, off for reproducibility.
• safety.mode - none/fast/full.
• outputFormat MP4/WEBM/MOV, outputQuality 20-99, numberResults 1-4.

Output is video (VIDEO).

How it works

Standard Runware pack plumbing: build a request with the model AIR and taskType, upload your frames, send over REST via the SDK, and get the video back as a native VIDEO. Because it's the fast tier, the round trip is shorter than the flagship node - which is exactly why you'd chain: image model → this node → preview → promote the good ones to the full 2.3 render. Cost shows on the title bar; the fast tier keeps that number small enough that iteration isn't painful.

Installing

ComfyUI Manager → search Runware → install → restart. Manual:

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

No model downloads; dependencies are runware-sdk, pillow, soundfile. API key via Settings → Runware API key, RUNWARE_API_KEY, or runware auth login.

Troubleshooting

Key first. If the node errors with nothing wired into frameImages, that's on you - it's required here by design, so check the socket. And remember this is the fast tier: if motion looks simplified on complex scenes, that's the speed trade, not a defect - promote the shot to the full Hailuo 2.3 node when it matters.