MiniMax Image to Video

If you've got a still image you want to move - a character, a product shot, a frame you generated and liked - this is the simplest path to video in ComfyUI's built-in partner category. Feed it one image and a prompt, and MiniMax's I2V-01 line animates that image forward as the first frame. No model loading, no sampler settings, no VRAM math. It's the "I just want a clip, please" node, and it delivers.

Like all the built-in partner/video nodes, this is an API node: it ships with ComfyUI core, uploads your image to Comfy Org's backend, calls MiniMax's video API for you, and charges your Comfy account in credits. The flat price badge reads about $0.43 a video regardless of what you prompt, which is unusual in this family - most of these nodes bill per second. That flat rate is both the appeal and the reason you should think before you spam it.

The inputs that matter

• image - your first frame. This is the anchor. Everything the video does starts here, and MiniMax's I2V is genuinely good at holding the subject's identity from the frame.
• prompt_text - what happens next. Describe the motion, not the contents of the image (the model can already see the image). "The camera slowly pushes in while she looks up and smiles" beats "a woman in a field."
• model - three tiers: the standard I2V-01, I2V-01-Director, and I2V-01-live. Director is MiniMax's prompt-heavier tier for more controlled, cinematic output; live is the cheaper, faster one. Default is plain I2V-01, which is honestly the right place to start.
• seed - an optional 64-bit seed. Treated loosely: it exists mostly so you can force a re-run for a different take.

Output is a single VIDEO object. Drop it into a save or preview node, or let the built-in UI show it.

What it's not

MiniMax's video models are closed - no weights, no local version. If you've been running Wan 2.2 locally and wondering why you'd ever pay for this, the honest answer is that MiniMax's I2V motion quality is a tier above what a 14B open model gives you without heavy tuning, and the price is predictable. The community lumps Hailuo and MiniMax into the "closed source" column of every big roundup, but that's exactly the trade these API nodes exist for: you keep ComfyUI's workflow graph and pay per video instead of buying a subscription elsewhere.

Where people get burned

• The "synchronous" description is optimistic. The node queues a task, shows a "Waiting/Processing" state while MiniMax renders, then downloads the result. Don't interrupt mid-run - the job is already charged.
• It's per-video pricing, so iteration is expensive. You can't dial quality down to save money the way you can with Kling's std mode or a local model. Draft and polish your prompt first; every render you don't do is money saved.
• The image is the contract. If your first frame is low-res or cluttered, the video inherits it. Clean up the frame before you wire it in.

For quick, decent, no-drama first-frame-to-video inside ComfyUI, this is one of the friendliest built-ins in the whole partner category. Just keep an eye on the running tab.