Aurora v1

Aurora v1 is Runware's hosted take on audio-driven image-to-video, from the creatify folks who build AI ad tools. You hand it a still image and an audio track and it hands back a video clip. No local GPU, no video diffusion checkpoints to babysit, no VRAM roulette - the whole thing runs on Runware's cloud and the node's title bar tells you what it cost you.

That last part is the real selling point. Local video gen is the single most annoying thing you can do in ComfyUI: multi-GB model downloads, low-VRAM crashes, minutes per clip. Aurora v1 sidesteps all of it. It's the node you reach for when a client or a project needs a talking segment or a sound-synced shot and you'd rather not spend the afternoon fighting Wan.

What you set

• image (required) - the starting frame, wired in as a plain IMAGE. Anything you'd feed a Video to Video workflow.
• audio (required) - the soundtrack, as a UUID or URL. Paste a hosted audio URL, or upload through Runware's media storage and drop the returned UUID in.
• positivePrompt - optional, and genuinely optional: it steers what the motion shows, but this model leans on the audio, not the prompt.
• CFGScale - defaults to 1. This model lives at low guidance; don't crank it.
• safety.mode - none / fast / full, a per-video content check. Off by default.
• outputFormat - MP4 / WEBM / MOV, and outputQuality (20–99) for how aggressively it compresses.

Output is video (VIDEO), which wires straight into Preview Video / Save Video. numberResults gives you up to 4 variations, each with a different seed.

How it works

Same mechanism as every node in this pack: it builds a request with the model AIR (creatify:aurora@0) and taskType: videoInference, ships it over REST through the Runware SDK, and pulls the rendered file back down as a native VIDEO object. The sync node code drives an async request under the hood, and the result shows up with a $0.00xxx cost tag on the node - that's your "am I burning money" readout while you iterate.

The one workflow gotcha: the audio must be reachable as a UUID or URL. There's no local-audio socket, so keep a hosted file or a media-storage UUID handy rather than trying to pass a path.

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, and soundfile (for the audio decode). You need a Runware API key: paste it into Settings → Runware API key, set RUNWARE_API_KEY, or run runware auth login once.

Troubleshooting

Missing-key errors are the default failure - the node won't run until a key is set. If the video comes back weird, check the obvious pair first: is the audio URL actually reachable, and did you keep CFGScale low? If the content check flags something and you're fine with it, that's what safety being off is for. And if you want the same job on a tighter budget, that's the Aurora v1 Fast sibling.