Wan2.6
Wan 2.6 without a 27B model on your desk
- frameImages
- referenceVideos
- video
Here's the thing about Wan 2.6 that most people don't realize until they go looking for the weights: it stopped shipping open. Wan 2.1 and 2.2 were the local community's default video models - Apache-licensed, everywhere, the reason "Wan" became shorthand for "decent open video" - but Alibaba moved 2.5, 2.6 and 2.7 behind their API. So this node is less "cloud convenience" and more "the only reasonable way to run Wan 2.6 from a ComfyUI graph." You get the model; you just don't get the gigabytes.
The inputs that matter
positivePrompt(required,STRING, multiline): what happens on screen. Wan is good at physics-y, lively motion, so describe action, not just subject.frameImages(optional,IMAGE): image-to-video. Feed a first frame (or frames) and Wan animates from it. This is the community favorite mode - anchoring the first frame solves a lot of consistency.providerSettings.alibaba.audio(defaulttrue): generate native audio aligned with the visuals. This is a Wan 2.x headline feature, and it's on by default - turn it off if you want the silent clip.providerSettings.alibaba.promptExtend(defaulttrue): LLM-based prompt rewriting. Here's the gotcha in the tooltip: it affects reproducibility. If you're chasing a specific seed, flip this off, or your "same prompt, same seed" won't be same at all.duration:5,10, or15seconds. 15 seconds at 1080p is a chunk of cloud compute, so budget accordingly.size: 720p/1080p presets in both orientations plus 1:1 and 4:3, and "from input" options that inherit your frame's aspect ratio.shotType:singleormulti- one continuous shot or a multi-scene composition.
The rest (negativePrompt, seed, numberResults up to 4, safety, ttl, outputFormat/outputQuality) is the standard Runware shared block. Output is a native video (VIDEO).
How it works
Pack-standard: your prompt and any frames go up as a REST request with model = Wan 2.6's AIR and taskType = video, Wan renders on Runware's GPUs, and the clip is downloaded and wrapped as a VIDEO for your graph. Cost shows on the title bar. Because there are no weights to manage, install is just the pack plus an API key.
Install and the gotchas
Install via ComfyUI Manager (search Runware) or clone https://github.com/Runware/ComfyUI-Runware into custom_nodes and pip install -r requirements.txt, then set a key (Settings → Runware API key, RUNWARE_API_KEY, or runware auth login).
Three things trip people up. Reproducibility: promptExtend on by default means your seed isn't your seed - turn it off for deterministic runs. Cost: 1080p/15s with audio and 4 results is a real bill; do your iterating at 720p/5s with numberResults at 1. Expectation: this is the API Wan, not the 2.2 you may have run locally - the model family moved on, and so did the price model. It's still the Wan you remember, just metered.
Inputs (19)
| Name | Type | Default | Description |
|---|---|---|---|
| positivePrompt | STRING | Text prompt describing elements to include in the generated output. | |
| frameImagesopt | IMAGE | — | |
| referenceVideosopt | RUNWARE_REFERENCEVIDEOS | — | |
| audioopt | STRING | Audio input (UUID or URL). | |
| negativePromptopt | STRING | Prompt to guide what to exclude from generation. Ignored when guidance is disabled (CFGScale ≤ 1). | |
| sizeopt | COMBO | 720p (16:9) | Output size. Only one of resolution or dimensions is sent. |
| seedopt | INT | 00–2147483647 | Random seed for reproducible generation. When not provided, a random seed is generated in the unsigned 32-bit range. |
| durationopt | COMBO | 5 | Length of the generated video in seconds. The total number of frames produced is determined by duration multiplied by the model's frame rate (fps). |
| numberResultsopt | INT | 11–4 | Number of results to generate. Each result uses a different seed, producing variations of the same parameters. |
| providerSettings.alibaba.audioopt | BOOLEAN | true | Generate native audio aligned with visual content. |
| providerSettings.alibaba.promptExtendopt | BOOLEAN | true | Enable LLM-based prompt rewriting to expand and clarify inputs. Affects reproducibility. |
| safetyopt | BOOLEAN | false | Enable to set safety. Off uses the model's default. |
| safety.checkContentopt | BOOLEAN | false | Enable or disable content safety checking. |
| safety.modeopt | COMBO | fast | Safety checking mode for video generation. |
| providerSettings.alibaba.shotTypeopt | COMBO | single | Shot composition style. |
| ttlopt | BOOLEAN | false | Enable to set ttl. Off uses the model's default. |
| ttl_valueopt | INT | 60 | Time-to-live (TTL) in seconds for generated content. Only applies when `outputType` is `URL`. |
| outputFormatopt | COMBO | MP4 | File format for the generated video. |
| outputQualityopt | INT | 9520–99 | Compression quality of the output. Higher values preserve quality but increase file size. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |