Kwaivgi Kling Video O3 Std First Last Frame (Channel Edition)
Cheap first-and-last-frame video on Kling O3, with a quiet catch
- first_frame_image
- last_frame_image
- api_config
- video
- request_id
- urls
The first-and-last-frame node is where Kling shows off: you pin two images and it invents a plausible, filmic way to get from one to the other. This is the Standard tier of that trick on the O3 line - the one you reach for when the endpoint frames matter more than the in-between quality, and the credits should stay reasonable.
What it is
A BizyAirPlus Kling channel node - cloud execution, no local weights, API key from bizyair.ai required. "Std" is O3's budget quality tier, and it's the honest default for FLF work: the interpolation between your frames is most of the creative content anyway, and Std handles it with only a modest quality drop against Pro. Where the O3 Std T2V sibling differs in one important way, the FLF version keeps the O3 multi-shot capability - so you can still script a multi-shot transition on the cheap tier.
How it works
Your first_frame_image anchors the start, the optional last_frame_image sets the destination, and the model generates the sequence between them. The frames are hard conditions - the model doesn't guess, it reconciles - which is exactly why mismatched endpoints produce visible contortions. O3's twist is that the bridge can be broken into multiple planned shots via multi_shot; on this node it defaults to off (note that - the Pro FLF node defaults it on), so you get a single continuous transition unless you opt in.
Inputs that matter
- first_frame_image - required. Where the shot starts.
- last_frame_image - optional. Where it has to end.
- prompt - required. Describe the movement and camera between the frames.
- duration - 3–15 seconds, default 5.
- sound - defaults to off here. Kling O3 can generate audio, but this tier's default is silent - flip it on if you want sound.
- multi_shot - default off; turn on for a multi-shot transition.
- multi_prompt / shot_type - the per-shot scripting controls, used when you enable multi-shot.
api_config and skip_error are the standard pack options.
Outputs
video → SaveVideo/preview. request_id is the cloud job id. urls are direct links to the rendered files.
Installing
ComfyUI Manager → search "BizyAirPlus", or:
cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt
Restart, toggle the BizyAirPlus button ON, enter your API key when prompted.
Where people get burned
The silent defaults, mostly. This node ships with sound off and multi_shot off, which is great for the budget but easy to misread as "the model can't do this." It can. Also the eternal FLF rule: your two frames need to agree on lighting, costume, and framing, or the model will visibly mutate the subject to bridge the gap. And while Std is the cheaper O3 tier, it's still metered cloud video - lock a seed and validate on short durations before you commit to a 15-second render.
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| first_frame_image | IMAGE | First image. | |
| prompt | STRING | Prompts. | |
| duration | INT | 53–15 | Video length in seconds. |
| sound | BOOLEAN | false | Whether to turn on the sound. |
| last_frame_imageopt | IMAGE | End frame picture, optional. | |
| multi_promptopt | STRING | Multi-shot prompt word configuration, JSON format. | |
| shot_typeopt | COMBO | intelligence | Type of shot. Customize is for customization and intelligence is for intelligence. |
| multi_shotopt | BOOLEAN | false | Whether to generate multi-shot video. |
| api_configopt | BIZYAIR_OPENAPI_CONFIG | 单次请求覆盖base_url和api_key | |
| skip_erroropt | BOOLEAN | false | 开启后遇到错误不中断工作流,输出对应类型的错误占位符 |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |
| request_id | STRING | — |
| urls | STRING | — |