wan / v2.2-5b / fal / text-to-video

Wan 2.2's 5B model was the "runs on an 8GB card" hero of the open-video world - and this node gives you that exact model without any card at all. It's text-to-video on fal's hosted fleet, routed through Civitai's Orchestration API as part of the official civitai-comfy-nodes pack. Prompt in, native VIDEO out, billed in Buzz.

The 5B is the sensible middle of the Wan 2.2 family: less jaw-dropping than the 27B MoE, but dramatically cheaper and faster to run, which is exactly what you want from a cloud node where every second and pixel is metered. This is the workhorse text-to-video node of the pack - the one you reach for when you want Wan 2.2 quality and motion without the bill that 1080p engines like Sora or Veo run up.

The inputs that matter

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

• resolution - 480p, 580p, or 720p (default 720p). 720p is this model's comfort zone.
• aspect_ratio - 1:1, 16:9, 9:16, or auto.
• num_inference_steps - 2 to 50, default 20.
• use_distill / use_fast_wan - the two speed-model toggles, both off by default. They cut turnaround and cost at the price of motion detail. The community's read on Wan 2.2 speedups is genuinely mixed - great for drafts, flattening for finals.
• interpolator_model - none, film, or rife (default film). Adds frame interpolation for smoother motion.
• negative_prompt - steer away from artifacts.
• enable_prompt_expansion - off by default; fal's server-side prompt rewrite. On for short prompts, off when your language is exact.
• enable_safety_checker - off by default; fal's safety pass.
• loras - the CIVITAI_LORAS socket. Trigger words go in prompt; the selector's keywords field is just a reminder.
• source_image also exists here (as on all the Wan fal nodes), 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.

Outputs

Standard pack 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, or sidebar sign-in.

Troubleshooting and takes

• Same inputs as the i2v twin. CivitaiVideoGenWanV225bFalImageToVideo is this node plus a first-frame input. If you're about to spend a credit on text-to-video, ask whether anchoring a frame would serve you better - I2V is where Wan's consistency wins.
• Text-only prompts drift. Without an anchor image, Wan 2.2 leans on the prompt entirely - describe motion and camera explicitly, or the clip will be a static scene that "moves."
• Fast modes flatten emotion. That's the 2.2 speedup tradeoff in one sentence. Draft fast, final slow.
• Slow / timeout. The pack shipped June 2026 during Civitai's GPU crunch; queues were long and timeouts happened. Default timeout is 30 min - raise via Civitai Auth timeout_minutes or CIVITAI_COMFY_TIMEOUT.
• Early preview. Nodes are generated from the API spec and "may change without notice."

Reach for this when you want the open-Wan 2.2 experience as a pure text-to-video cloud call with a modest meter. It's the budget-quality sweet spot of this pack's video nodes.