Kwaivgi Kling V3.0 Pro First Last Frame (Channel Edition)
Kling 3.0 Pro first-and-last-frame video, channel edition
- images
- api_config
- video
- request_id
- urls
Kling 3.0 Pro - the current quality benchmark from Kuaishou's AI lab (the KwaiVGI people behind LivePortrait and the whole Kling line) - in first-last-frame mode, served through SiliconFlow's BizyAir channel. FLF2V means you hand the model two images: how the video starts and how it ends. The model invents everything in between. That's a step up in control from plain image-to-video, and it's the mode people actually use when they need a scene to land somewhere specific.
You reach for this when "animate this image" isn't enough and you need a destination. Character transformation shots, scene transitions, storyboard end-frames - FLF gives you a defined narrative arc instead of hoping the model wanders somewhere good. For the price of preparing two stills, you get a video whose ending you already approved.
How it works
BizyTRD cloud wrapper. images is an autogrowing input (drop in your first and last frames), and when you upload a first frame the system auto-adapts to the image's ratio - the tooltip says aspect_ratio is optional for that reason. duration is an integer 3–15 seconds (default 5). prompt is optional but you'll want it. seed defaults to 0 with -1 for auto, and there's a sound toggle (default on) for generating sound effects with the clip. api_config overrides endpoint/key per request; skip_error outputs a placeholder instead of aborting the workflow.
Inputs and outputs
images(required, autogrowing) - first and last frames.duration(required, 3–15s, default 5).- Optional:
prompt,aspect_ratio(default9:16),seed,sound,api_config,skip_error. - Outputs:
video(VIDEO),request_id,urls.
Installing BizyAirPlus
ComfyUI Manager: "BizyAirPlus", or:
cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAirPlus.git
cd BizyAirPlus
python -m pip install -r requirements.txt
Restart, account at bizyair.ai, API key, BizyAirPlus button ON. Dependencies: bizyair-cloudberry, bizytrd - no models to download.
Common issues
Two things trip people up. First, frame order: the autogrowing images input takes them in order, and swapping first and last turns your narrative into a reverse playthrough - check the wire order. Second, the sound toggle: it defaults on, which is delightful until you realize every clip is billed with audio generation and you didn't need it. Turn it off for silent b-roll. Then the pack standard - missing button (pip show bizyair-cloudberry), cloud errors in the console, metered pricing. Kling Pro is one of the pricier tiers in this pack, so it's worth pinning a good seed and reusing it rather than gambling per generation.
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| images | COMFY_AUTOGROW_V3 | — | |
| duration | INT | 53–15 | Video length, in seconds, range 3-15 seconds |
| promptopt | STRING | Prompt words for video generation | |
| aspect_ratioopt | COMBO | 9:16 | Aspect ratio of video. When uploading the first frame, the system will automatically adapt the image ratio, this parameter is optional. |
| seedopt | INT | 0-1–2147483647 | Random seed, -1 means automatically generated |
| soundopt | BOOLEAN | true | Whether to turn on sound effects |
| 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 | — |