wan / v2.2 / fal / image-to-video

Wan 2.2 is the model that ended the "which local video model?" argument - and this node serves it up on fal's fleet through Civitai's cloud, as part of the official civitai-comfy-nodes pack. First frame in, clip out as a native VIDEO socket, billed in Buzz. No checkpoint download, no VRAM ceiling, no local ComfyUI for the generation part at all.

Image-to-video is where 2.2 shines. Anchoring to a real frame is the difference between Wan 2.2's "more motion, more emotion" output actually being usable and being a dice roll. If you have a still you like, this is the node that makes it move.

The inputs that matter

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

• source_image / images - your starting frame. The heart of the i2v node.
• resolution - 480p or 720p (default 720p). 2.2's fal tier tops out at 720p here; 1080p lives on the closed 2.5/2.6 nodes.
• aspect_ratio - seven options: 1:1, 16:9, 9:16, 4:3, 3:4, 4:5, 5:4 (default 16:9).
• shift - 1.0 to 10.0, default 5.0. Wan's training-shift dial; controls denoising focus. Tune it only if you know what you're doing - defaults are sane.
• use_turbo - off by default. fal's turbo path: faster and cheaper, flatter output. Draft with it, final without.
• interpolator_model - film or rife (default film). Frame interpolation for smoother motion between generated frames.
• negative_prompt - steering; useful with CFG control.
• enable_prompt_expansion - off by default; fal's prompt rewrite.
• enable_safety_checker - off by default; fal's safety pass.
• loras - CIVITAI_LORAS socket; trigger words go in prompt, the selector's keywords field is just a reminder.
• 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 sidebar tracking, 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

• 2.2's quality ceiling is real, and so is its slowness. Locally, a 720p clip took ~25 minutes on a 5090 without speed LoRAs. On fal's cloud you rent that speed, but it shows up as queue time and a meter - pack's launch (June 2026) coincided with Civitai's GPU crunch, so long waits and timeouts were the top complaints. Default timeout 30 min; raise via Civitai Auth timeout_minutes or CIVITAI_COMFY_TIMEOUT.
• Turbo vs quality is a real trade. use_turbo softens the motion detail that makes 2.2 special. The community's compromise: turbo for drafts, full quality for the keeper.
• Text-only prompt drift? You're on the i2v node - use a frame. If you have no image, the text-to-video sibling is the same recipe minus the anchor.
• Negative prompt + safety checker interact. With enable_safety_checker on, fal's own pass runs before output; that's independent of your negative prompt.
• Early preview. Generated nodes "may change without notice."

Reach for this when you want the best open-weights motion quality in the pack applied to a frame you already like. If you want to drive the sampler yourself instead of letting fal's defaults run, the "comfy" Wan 2.2 node is the same model with the full control surface.