wan / v2.5 / fal / text-to-video

Wan 2.5 never got open weights - Alibaba previewed it as a cloud model in September 2025 and that was the end of it. So this node, running 2.5 text-to-video on fal's fleet through Civitai's Orchestration API, is one of the few legitimate doors into the model. Prompt in, native VIDEO out, billed in Buzz. It's part of the official civitai-comfy-nodes pack.

The clean pitch: Wan 2.5 text-to-video at 1080p, which the open 2.1/2.2 fal nodes don't offer, with a much simpler control surface than the 2.2 "comfy" node. The honest caveats: 2.5's quality bump over 2.2 was modest, it's a closed model with a server-side filter you can't patch around, and every run is metered.

The inputs that matter

prompt and cfg_scale (default 4) are required. Then:

• resolution - 480p, 720p, or 1080p (default 1080p). The reason this node exists over the 2.2 ones.
• aspect_ratio - 16:9, 9:16, or 1:1 (default 16:9).
• enable_prompt_expansion - on by default. 2.5's fal tier expects you to lean on prompt expansion; flip it off only when your language is exact and you want it verbatim.
• negative_prompt - steering.
• loras - CIVITAI_LORAS socket; trigger words go in prompt, not the selector's keywords field.
• source_image also exists (all the Wan fal nodes carry it), but for pure text-to-video you can leave it empty.
• duration 1–30s (default 5), frame_rate 24, steps 10–50, seed, api_config.

That's the whole surface - no shift, no turbo toggle, no interpolator. Simpler than 2.2's fal nodes, which is either a relief or a loss depending on who you are.

Outputs

Standard trio: video (VIDEO) into a Save Video / preview node, workflow_id (STRING) for the Civitai sidebar, raw_json (STRING) for response and cost.

Install and auth

One-time, pack-wide. ComfyUI Manager → search Civitai Comfy Nodes → install, restart:

cd ComfyUI/custom_nodes
git clone https://github.com/civitai/civitai-comfy-nodes.git
pip install -r civitai-comfy-nodes/requirements.txt   # just `requests`

Auth: Civitai Auth node, CIVITAI_API_TOKEN (headless), or sidebar sign-in.

Troubleshooting and takes

• "Is this better than Wan 2.2?" Marginal, and the community's honest answer is "probably not worth the premium if you don't need 1080p." The open 2.2 nodes in this pack are cheaper, uncensored-by-default, and better understood. 2.5's case is resolution and the newest closed polish.
• Prompt expansion is on by default. Output drifting from your text? That's it. Turn it off and compare.
• Closed model, fixed filter. No local patch exists for 2.5's moderation. If a prompt refuses here, it's Alibaba's call, not a broken install.
• Slow / timeout. The pack shipped June 2026 into Civitai's GPU crunch - long queues, occasional timeouts, and 1080p makes them likelier. Default timeout 30 min; raise via Civitai Auth timeout_minutes or CIVITAI_COMFY_TIMEOUT.
• Early preview. Generated nodes "may change without notice"; re-check fields after updates.

Reach for this when you want the closed-Wan generation at 1080p with minimal dials. Have a first frame to anchor? The image-to-video sibling gets you the consistency that pure text-to-video can't.