⚠️ [Deprecated] Generate Video (from Text)

DeepGen_T2V is the deprecated name of the pack's text-to-video node. Load an old workflow and you'll see it as "⚠️ [Deprecated] Generate Video (from Text)" - and as with every deprecated node in this pack, the warning is about the name, not the node. It's the same class that now ships as DeepGen_T2V0, kept as a compatibility alias since the 1.2.1-era rename. Same inputs, same outputs, same cloud calls; the only differences are the display name, the DeepGen/Deprecated category, and the flag.

What it does

Type a prompt, pick a video model, and a cloud service renders a clip that comes back into ComfyUI as a VIDEO type. The model list is the serious closed-video roster: Kling 2.5 Turbo / 2.5 Turbo Pro, Kling 3.0 and 3.0 Pro / Omni, Veo 3.1 and 3.1 Fast, Seedance 1.5 Pro, Vidu Q3 Pro, LTX-2.3 Fast / Pro, Grok Imagine Video. These are exactly the models you cannot run locally - Veo's native-audio generation, for instance, is a capability with no open-weight equal (per the KB's closed-model essay) - and this node is how they enter a ComfyUI graph.

Mechanically, video is the pack's slow path: the node submits the job with queue: true, then polls every 15 seconds until the clip is done, which can mean minutes of waiting with your queue blocked. When the job completes, the video is downloaded to ComfyUI's temp directory and handed back as a ComfyVideoMock object - the pack's stand-in for ComfyUI's native VIDEO type - which preview and save-video nodes know how to handle.

The inputs that matter

• model - the video-model dropdown.
• prompt - what the clip should show. Video models reward structured, cinematic language.
• duration (1–15, default 5) - clip length in seconds; the pack snaps it to what the model actually supports. This is your main cost dial.
• generate_audio (default on) - Veo and the Kling Omni models can add audio; turning it off saves credits.
• aspect_ratio and resolution - free-form strings (leave empty to let the model choose, or type e.g. 16:9 / 720p).
• nb_results, seed_value, output_prefix, config_json - the standard set.

Outputs

VIDEO (the clip, for preview/save), output_prefix_and_model (STRING), and total_credits_used (FLOAT). Video is where credit burn gets real - a multi-second 1080p clip costs multiples of a still image, and the KB's API-node essay quotes the community reaction to exactly this reality.

Install

Same pack:

cd ComfyUI/custom_nodes
git clone https://github.com/deepiksdev/ComfyUI-DeepGen-API.git
cd ComfyUI-DeepGen-API
pip install -r requirements.txt

Restart, then set your DeepGen API key (deepgen.app) in Settings → "DeepGen API Key" (ComfyUI/user/deepgen/config.json).

Gotchas

Budget for the wait - the node blocks while polling, so don't put ten of these in a queue and walk away. Watch the cost, especially with generate_audio on. And for new work, use DeepGen_T2V0; this name is the legacy spelling. If your old graph references DeepGen_T2V, it'll keep running as-is.