NS WaveSpeed Wan 2.5 Text to Video
The quality jump you can't download
- client
- video_url
Wan 2.2 made Alibaba's video models the community's default; Wan 2.5 is what people wanted next, and then it wasn't there. No open weights, no local install, no argument about VRAM - 2.5 and everything after it is API-only (wan-video.md). NSWaveSpeedWan25TextToVideo from symbiotica-ai/comfyui-nodes is the ComfyUI door to it: prompt in, hosted render on WaveSpeed, video_url out. WaveSpeed was the first platform to host Wan 2.5, so this is close to the original pipe.
How it works
Prompt plus a size goes to WaveSpeed's Wan 2.5 t2v endpoint, and you get a video_url string back. Six sizes are on the menu, and this is where the node earns its keep: 832*480, 480*832, 1280*720, 720*1280, 1920*1080, and 1080*1920. Real 1080p, both orientations - that's the resolution tier that separates 2.5 from the 2.2 nodes in this same pack.
The extras make it feel like a proper video-generation tool rather than a thin wrapper. audio takes a URL to a wav/mp3 (3–30 seconds, ≤15MB) that guides the video - an audio-conditioned generation, not just a silent clip. duration is 5 or 10 seconds. enable_prompt_expansion (off by default) lets Wan enrich your prompt before rendering, which is genuinely useful for a bare phrase like "a fox crossing a snowy field." And there's a negative_prompt - worth filling in, same as the 2.2 nodes.
enable_sync_mode defaults off: the node submits, polls every second, and blocks the graph until the clip's done - with Cancel properly honored during the poll.
The inputs that matter
prompt- the scene, the motion, the vibe.size- six options, up to 1920x1080 / 1080x1920.audio(optional) - a URL to audio that conditions the video.duration- 5 or 10 seconds.enable_prompt_expansion- auto-enrich the prompt.client- theWAVESPEED_AI_API_CLIENTsocket from the NS WaveSpeed Client node.
Output: video_url (STRING) - preview in a browser or feed a download-from-URL node.
Install and keys
# ComfyUI Manager: search "Symbiotica", or:
cd ComfyUI/custom_nodes
git clone https://github.com/symbiotica-ai/comfyui-nodes.git symbiotica
pip install -r symbiotica/requirements.txt
Restart, add the NS WaveSpeed Client node with your wavespeed.ai key (widget, Settings → Symbiotica, or WAVESPEED_API_KEY), wire client in. No model files.
Common issues
The local-Wan reflex is wrong here. Everything you know about downloading Wan stops at 2.2. If you want 2.5 specifically, hosted is the only lane - and the community verdict on 2.5 over 2.2 is genuinely positive, so the price may be worth it for the 1080p and the better prompt adherence (wan-video.md).
Audio-conditioning adds a constraint. The audio URL has to be 3–30 seconds and ≤15MB, and it has to be hosted somewhere WaveSpeed can fetch. If the render fails oddly and you've wired an audio input, check that first.
Billing, waiting, moderation. Minutes per render, per-call pricing, and your prompt leaves the machine. The standing contract for every API node in this category (external-api-nodes.md).
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| client | WAVESPEED_AI_API_CLIENT | — | |
| prompt | STRING | Text description for video generation | |
| size | COMBO | 1280*720 | Video resolution (width*height) |
| negative_promptopt | STRING | Describe what you don't want in the video | |
| audioopt | STRING | Audio URL to guide video generation (3-30 seconds, wav/mp3, ≤15MB) | |
| durationopt | COMBO | 5 | Video duration in seconds |
| enable_prompt_expansionopt | BOOLEAN | false | Automatically expand and enhance the prompt |
| seedopt | INT | -1-1–2147483647 | Random seed for reproducible results. -1 for random seed |
| enable_sync_modeopt | BOOLEAN | false | Wait for generation to complete before returning |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| video_url | STRING | — |