Generate Video (from 7 Images)

DeepGen_I2V7 is the pack's maximum-keyframe video node: seven images in, one clip out, with the model expected to pass through every frame in order. It's the closest thing this pack has to a storyboard-to-video node - if you've ever drawn up a seven-panel sequence and wished it would just animate itself, this is that.

Why you'd reach for it

Seven frames is real direction. With three you get a beginning, middle, end; with seven you get beats - an establishing shot, an action, a reaction, a wide, a close-up - and the model has to honor all of them. It's the tool for anyone who works out a scene as a sequence of panels before generating, which is a genuinely good habit with video models: the more you constrain the structure, the less the model improvises something you didn't ask for. The tradeoff is the model list, which is the shortest in the video family - only Kling 3.0 Omni Pro, Kling 3.0 Omni Standard, and Grok Imagine Video can hold seven frames of context. If you need a model that isn't listed, it doesn't do seven-frame input, and no setting will change that.

How it works

The seven images become base64 PNG attachments submitted with the prompt; the job queues and the node polls every 15 seconds until the clip renders - minutes of waiting with your workflow blocked. When done, the video downloads to ComfyUI's temp dir and returns as ComfyVideoMock, the pack's VIDEO-type stand-in, ready for the usual preview and save nodes.

The inputs that matter

• image_1 through image_7 (IMAGE) - the frames, in sequence. All optional but the node only makes sense with most or all connected. This is a directed sequence, not a suggestion bag - order is 1 → 2 → … → 7.
• prompt - the connective tissue: the action, camera, and tone that link the panels.
• model - the seven-frame-capable trio.
• duration (1–15, default 5), generate_audio (default on), aspect_ratio / resolution, plus nb_results, seed_value, output_prefix, config_json - the standard video set; duration and resolution move the cost.

Outputs

VIDEO (the clip), output_prefix_and_model (STRING), total_credits_used (FLOAT).

Install

Same pack, standard steps:

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

Restart and add your DeepGen API key (deepgen.app) in Settings → "DeepGen API Key", stored at ComfyUI/user/deepgen/config.json.

Gotchas

The more keyframes you feed, the harder the reconciliation problem - seven frames that contradict each other will come out as a compromise video. Keep the subject and style consistent across panels, and treat the prompt as the glue, not the star. This is also the most expensive image-to-video mode per call, so test with short durations and generate_audio off before committing to full renders.