Kling AI Multi-Shot
One prompt, six shots — consistent characters across a storyboard
- auth
- images
- url
- task_id
The pain this node kills: generating a storyboard and having the character look like a different person in every panel. Kling's multi-shot image mode tackles that head-on by generating up to six shots as one task, with a shared character/scene description anchoring consistency across all of them. Instead of six separate generations and a prayer, you get a batch where the same character stays recognizably the same person. If you storyboard, design a character sheet, or pitch sequences, this is the node you'll actually reach for.
Mechanically it's a single cloud task with multiple shots: the required prompt describes the overall scene and character (the consistency anchor), shot_1_prompt is the first shot (required), and shot_2_prompt through shot_6_prompt are optional - leave them blank and you get fewer shots. One aspect_ratio applies to all of them. The result comes back as a batch of images, so the images output is an IMAGE tensor you can preview as a grid or split per-frame.
The inputs that matter:
prompt- the character/scene anchor. This is where you lock in "the detective, brown trench coat, neon-lit city" so it applies everywhere.shot_1_promptthroughshot_6_prompt- what happens in each panel. Consistency comes from the anchor prompt, so keep it descriptive and don't fight it with conflicting details in the shot prompts.aspect_ratio- same for every shot.
Outputs: images (batch), url, task_id.
Installing it
Bundled in ComfyUI-API-Toolkit. Manager: search "API Toolkit". Manual:
cd ComfyUI/custom_nodes
git clone https://github.com/IxMxAMAR/ComfyUI-API-Toolkit
pip install -r requirements.txt
Restart ComfyUI. Needs the Kling AI Authentication node and a funded account.
Gotchas
Consistency is much better than separate generations, but it's not clone-level. Kling holds the character, then drifts under heavy motion or big angle changes across shots - if your storyboard demands the exact same face across all six panels, consider Kling's element/character system (the pack's Element Selector) as a companion, or fix faces in a local pass afterward. Also watch the credit meter: this is one task that generates multiple images, and multi-image tasks bill accordingly. Draft with two or three shots, then expand to six once the character design is locked. And remember the whole batch shares one aspect ratio - a mix of portrait and landscape panels isn't possible in a single task.
Inputs (9)
| Name | Type | Default | Description |
|---|---|---|---|
| auth | KLING_AUTH | — | |
| prompt | STRING | Overall scene/character description for consistency across shots. | |
| shot_1_prompt | STRING | Prompt for shot 1 (required). | |
| aspect_ratio | COMBO | 1:1 | Aspect ratio for all generated shots. |
| shot_2_promptopt | STRING | Prompt for shot 2. | |
| shot_3_promptopt | STRING | Prompt for shot 3. | |
| shot_4_promptopt | STRING | Prompt for shot 4. | |
| shot_5_promptopt | STRING | Prompt for shot 5. | |
| shot_6_promptopt | STRING | Prompt for shot 6. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |
| url | STRING | — |
| task_id | STRING | — |