BYOKey Runway Text to Image
Runway makes images now, and you can drive Gen4 with your own key
- reference_image
- output
You probably think of Runway as a video company, but Gen4 does still images too, and they're good - cinematic in a way that matches the video side. This node is the bring-your-own-key way to get them into a ComfyUI graph. The nice part: pair it with BYOKey_Runway_Image2Video in the same pack and you've got a fully Runway pipeline - image out of this node, straight into the video node as its start frame.
How it works. One POST to api.dev.runwayml.com/v1 (the default base_url, overridable for a relay) with your key as a Bearer token. Runway's image endpoint is async, so the node polls until the image is done (default every 5 seconds, up to 5 minutes), then decodes the PNG into a [1, H, W, C] torch tensor for standard image preview/save nodes. An optional reference_image gets sent along to guide generation - Runway's Gen4 is genuinely good at carrying a reference's look into a new image.
The inputs that matter:
prompt- your text description. Cinematic phrasing plays to Gen4's strengths; think in terms of lens, light, and frame.ratio- 10 presets from1920:1080down to portrait1080:1920, including the slightly unusual2112:912and1808:768ultrawides. Pick your format before you queue.reference_image- optionalIMAGEinput. Feed it a character, a style, or an existing render to steer the result. This is the input that makes the node more interesting than a plain t2i call.poll_interval/max_wait- the usual babysitters. Image jobs are fast, so the 300s default cap rarely gets hit.
One thing you won't find here: no seed input. Runway's image API doesn't expose one through this node, so don't go hunting for reproducibility - you get what you get and re-roll if you don't like it.
Output: one output (IMAGE) socket. Straight into Save Image, or into the Runway i2v node next door.
Install. Shared pack, one clone:
cd ComfyUI/custom_nodes
git clone https://github.com/MeteorAndy/comfyui-byokey.git
# restart ComfyUI; it's under api/byokey/image
No model downloads, no extra dependencies beyond ComfyUI's bundled torch, aiohttp, PIL, numpy and av.
Gotchas. No seed means every run is a fresh roll - fine for exploration, annoying if you wanted a specific look back. And the pack is young with zero community footprint, so the standing advice applies: skim the source before pasting a real key in.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| api_key | STRING | Runway API key (sent as Bearer token). | |
| base_url | STRING | https://api.dev.runwayml.com/v1 | Runway API base URL. Override for relays. |
| prompt | STRING | Text prompt for the generation. | |
| ratio | COMBO | 1024:1024 | Aspect ratio for the generated image. |
| reference_imageopt | IMAGE | Optional reference image to guide the generation. | |
| poll_intervalopt | INT | 51–120 | Seconds between status polls. |
| max_waitopt | INT | 30030–3600 | Max seconds to wait for the task. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| output | IMAGE | — |