Happy Horse / v1.1 / textToVideo

The v1.1 text-to-video node from the Happy Horse family - prompt goes in, a clip comes back, and the work happens on Civitai's cloud through the official civitai/civitai-comfy-nodes pack. No GPU needed, no model to download, and every run is billed in Buzz.

If you're choosing between the v1.0 and v1.1 text-to-video nodes, the honest answer is: v1.1 is the newer engine Civitai is actively shipping, and it's the one with the wider framing options. The aspect_ratio dropdown here has nine entries (16:9, 9:16, 1:1, 4:3, 3:4, 21:9, 9:21, 5:4, 4:5) against v1.0's five, which matters if you want ultrawide cinematic framing or near-square output straight from the node instead of a crop. There's no strong community lore yet on the exact quality delta between the two versions - it's a hosted model, so you can't inspect the weights - but you'd normally reach for v1.1 unless you specifically need the older behavior.

How it works

Identical orchestration machinery to the rest of the pack: build a videoGen workflow discriminated to happyHorse / v1.1 / textToVideo, submit to the Civitai Orchestration API, long-poll until the job succeeds, then download the MP4 into ComfyUI's temp directory and return it as a native VIDEO. Your prompt goes straight to Civitai's fleet. That's the mechanism in full - there's nothing running on your side beyond the HTTP client, which is exactly why the node installs so cleanly and runs on any machine ComfyUI runs on.

Inputs and outputs that matter

• prompt (required) - the whole clip in words: subject, action, camera move, lighting, style. Cloud video models are as hungry for detail as local ones.
• resolution - 720p / 1080p, default 1080p.
• duration - 3–15 seconds, default 5.
• seed - 0 for random; set to lock a result.
• aspect_ratio - the full nine, described above.
• api_config - optional Civitai Auth connection; defaults to CIVITAI_API_TOKEN or stored OAuth.

Outputs: video (the clip), workflow_id, raw_json (the raw job response, useful when debugging failures).

Installing it

One pack install covers it. ComfyUI Manager → Custom Nodes Manager → "Civitai Comfy Nodes" (publisher civitai) → Install → restart, or manually:

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

Only requests is needed; there are no model files.

Common issues

• Credentials. Civitai account + Buzz required. On headless installs set CIVITAI_API_TOKEN; otherwise the Civitai Auth node or sidebar connect panel works.
• Cost. Cloud video generation is metered and adds up. The node's status line shows workflow id and cost per run - worth checking before you batch fifty seeds.
• Prompt quality is on you. The model follows what you write; vague prompts give vague clips. Write camera and motion explicitly.
• Early preview. The pack is under active development; APIs and node shapes can change without notice. Pin the version for anything you rely on.
• Inputs leave the machine. Prompt text is sent to Civitai's servers - standard for this whole category, but worth remembering.