flux / v3.0 / textToVideo
Type a sentence, get a video clip
- api_config
- video
- draft_cache_audio
- draft_cache_video
- workflow_id
- raw_json
Of the three FLUX 3 video nodes in this pack, this is the one with the least to trip over: it has exactly one required input - prompt - and no image, no JSON, no keyframes. You type a sentence about what you want moving, and a video comes back. If you're new to the whole civitai-comfy-nodes pack, this is the node to learn on before you touch the image or keyframe variants.
It's part of Civitai's official pack of ~160 nodes, all generated from the Orchestration API spec. The node is a thin client for FLUX 3 Video, Black Forest Labs' closed multimodal model (image + video + audio in one set of weights) - the weights aren't downloadable, so Civitai runs them in the cloud and you pay in Buzz. Under the hood the node builds a videoGen job with the flux / v3.0 / textToVideo discriminator, submits it, long-polls Civitai's fleet until it's done, and hands you a native VIDEO tensor. To ComfyUI, it's just another generator node sitting in your graph.
The dials that matter
Everything beyond prompt is optional, which is the whole appeal:
- duration - 5 to 20 seconds. Every second is billed, so the default 5 is a fine starting point.
- resolution - 720p or 1080p.
- aspect_ratio -
autoby default, or pin it: 16:9, 9:16, 1:1, 21:9, 4:3, 3:4, 2:1. - generate_audio - on by default. FLUX 3 generates native synchronized audio (dialogue, SFX, ambience) with the video; flip it off if you're doing your own sound.
- draft - renders a fast, cheap 720p draft instead of the full-quality render. Draft results carry a draft cache that a later
draftEnhancestep re-renders at full quality - the "sneak preview before you pay full price" switch.
Outputs are video (VIDEO), draft_cache_audio (AUDIO), draft_cache_video (VIDEO), and the two debug strings, workflow_id and raw_json. The node stamps its workflow id and cost onto itself after each run, which is how you keep an eye on what a session is costing you.
Install and auth
Standard for this pack - official, on the Comfy Registry, one dependency file:
# ComfyUI Manager: search "Civitai Comfy Nodes"
cd ComfyUI/custom_nodes
git clone https://github.com/civitai/civitai-comfy-nodes.git
pip install -r civitai-comfy-nodes/requirements.txt
Restart ComfyUI, then authenticate via a Civitai Auth node, the CIVITAI_API_TOKEN env var, or the sidebar OAuth login. Headless/remote installs should use the env var - the README calls that out explicitly.
The honest gotchas
Cost is the big one. Text-to-video is metered per generated second, and FLUX 3's strong prompt adherence means you'll want retries - which are also billed. Draft mode is the pressure-relief valve. Your prompt leaves your machine, full stop - that's inherent to any API-wrapper node and this one is no exception. And the pack is early preview: nodes, APIs and behavior "may change without notice" per the README. Early evaluations also flagged FLUX 3 Video struggling with consistency across multiple shots and complex actions - fine for a single clip, less reliable for anything multi-scene. For that, the keyframes variant exists, but honestly, start here.
Inputs (7)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | — | |
| durationopt | INT | 55–20 | Duration of the generated video in seconds. |
| resolutionopt | COMBO | 720p | 2 options: 720p, 1080p |
| aspect_ratioopt | COMBO | auto | Aspect ratio of the generated video. "auto" lets the model choose. |
| generate_audioopt | BOOLEAN | true | Whether to generate a synchronized audio track for the video. |
| draftopt | BOOLEAN | false | Render a fast, cheap 720p draft instead of a full-quality video. Draft results additionally carry a draft cache that a later draftEnhance operation re-renders at full quality. |
| api_configopt | CIVITAI_CONFIG | Optional Civitai Auth connection; defaults to CIVITAI_API_TOKEN or stored OAuth login. |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |
| draft_cache_audio | AUDIO | — |
| draft_cache_video | VIDEO | — |
| workflow_id | STRING | — |
| raw_json | STRING | — |