haiper

Of all the cloud video nodes in Civitai's pack, this is the one with the most knobs. CivitaiVideoGenHaiper runs Haiper's video model on Civitai's cloud with everything exposed: negative prompt, camera movement presets, seed, duration, aspect ratio, model version (v1_5 or v2), resolution up to 2160p, an AI prompt-enhancer toggle, and an optional source image for image-to-video. If you're the type who likes control over your generators, this is your node. It costs Buzz per clip, and Haiper's weights aren't downloadable, so the cloud is the only way in.

It lives in Civitai/Video/haiper, part of Civitai's official ComfyUI pack. Where the grok nodes are minimal (prompt, duration, resolution), Haiper gives you the full production kit. That camera-movement dropdown - panRight, panLeft, tiltUp, tiltDown, zoomIn, zoomOut - is the standout: it's the difference between "a video of a street" and "a slow zoom into a street at dusk."

How it works

The node submits a videoGen workflow (engine: haiper) to Civitai's Orchestration API. Your prompt, settings, and any source image go up; the cloud runs Haiper; the pack long-polls and returns a native VIDEO tensor plus workflow_id and raw_json. When enable_prompt_enhancer is on (default), the model rewrites your prompt internally before generating - leave it on unless it's over-stylizing your output.

The inputs that matter

• prompt (required) - what you want on screen.
• negative_prompt - what you don't. Useful for killing common artifacts ("blurry, warped hands, jitter").
• camera_movement - panRight, panLeft, tiltUp, tiltDown, zoomIn, zoomOut, or none. This single dial shapes the feel of the clip more than anything else.
• source_image - wire in an image for image-to-video; leave empty for pure text-to-video.
• model - v1_5 or v2. Default empty (let the API choose); if you've tested both, pin your favorite.
• duration - 2, 4, or 8 seconds.
• resolution - 720, 1080, or 2160. 2160 is the money shot and the money bill.
• aspect_ratio - 16:9, 9:16, 4:3, 1:1, 3:4.
• seed - set it for reproducible shots; leave it to get variety.
• enable_prompt_enhancer - default on; the AI rewrites your prompt for better results.

Installing it

This is one of ~160 nodes in Civitai Comfy Nodes, Civitai's official pack for their Orchestration API:

• ComfyUI Manager: Manager → Custom Nodes Manager → search Civitai Comfy Nodes → Install, then restart.
• CLI: comfy node registry-install civitai-comfy-nodes
• Source: cd ComfyUI/custom_nodes && git clone https://github.com/civitai/civitai-comfy-nodes.git && pip install -r civitai-comfy-nodes/requirements.txt (just requests).

You need a Civitai account with Buzz and credentials - a Civitai Auth node, CIVITAI_API_TOKEN (best for headless), or a stored key via the Civitai sidebar.

Where people get burned

• No credentials → hard error. Export CIVITAI_API_TOKEN and restart ComfyUI.
• Resolution is a direct cost lever. 2160p looks great and bills like it. Run tests at 720p and only spend up for finals.
• Inputs leave the machine. Prompt and any source image go to Civitai's servers and Haiper's model, with both parties' moderation. Closed model, unfixable filter.
• Video is the expensive meter. Per-second cloud video adds up - the standing community complaint about API video nodes applies fully here. Check the node's cost report after each run.
• Early preview. The README warns of breaking changes without notice; early community reports flagged slow jobs and bugs. Long or high-res clips can hit the default 30-minute timeout - raise it via the Auth node or CIVITAI_COMFY_TIMEOUT.

Haiper is the node you go to when the minimal generators aren't enough - when you need a specific camera move, a negative prompt, and a real resolution choice. It's the most "production" of the cloud video nodes here, and priced like it.