Kling 2.6 Image(First Frame) to Video with Audio
Kling 2.6 first-frame to video that ships with sound baked in
- start_frame
- VIDEO
The Kling Image to Video with Audio node (Kling 2.6) does what its name says on the tin: give it a first-frame image, get back a video that already has sound. That's the trick that makes it worth reaching for. Most video generators give you moving pixels and leave you to score the audio yourself; this one hands you a clip where the ambient sound, dialogue, and foley are already synced to what's on screen.
It's the image-to-video sibling of the Kling 2.6 text-to-video-with-audio node, and it runs through the same pipeline: the node uploads your image to Kling's servers via your Comfy account, generates the video with sound turned on or off depending on your toggle, polls until the job's done, and returns the finished VIDEO. No local model, no downloads - billed to Comfy account credits per generation, with audio roughly doubling the cost of a silent clip.
The inputs that matter:
- start_frame - your first frame. This is an
IMAGEinput, so wire in a Load Image node. Kling validates it: minimum 300×300px, aspect ratio between 1:2.5 and 2.5:1, and it gets uploaded rather than processed locally. - prompt - positive text prompt describing the motion and scene. 2,500 character cap.
- model_name - fixed to
kling-v2-6for this node. - mode -
prois the only option here. - duration - 5 or 10 seconds.
- generate_audio - the toggle that matters. On (default) means the output carries generated sound; off gets you silent footage and a cheaper bill.
The single VIDEO output feeds straight into Save Video or Preview Video.
The one thing to keep straight: this node is not the same as Kling 3.0. It's the 2.6-era model, and while 2.6 isn't facing the same retirement clock as the V1.x line, the Kling 3.0 nodes (KlingVideoNode, the Omni pair) are where the newer capabilities live. Use this one when you specifically want a first-frame, 2.6-style result with audio, or want to match clips generated on 2.6.
Practical gotchas: watch the credit burn if you leave generate_audio on for every pass - it's roughly twice the price of silent video, and you'll be re-rolling. And remember this is a partner/API node: you must be logged into your Comfy account or it fails immediately with an auth error. If the partner-node family is cluttering your node menu, --disable-api-nodes in your launch args removes them all.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| model_name | COMBO | 1 options: kling-v2-6 | |
| start_frame | IMAGE | — | |
| prompt | STRING | Positive text prompt. | |
| mode | COMBO | 1 options: pro | |
| duration | COMBO | 2 options: 5, 10 | |
| generate_audio | BOOLEAN | true | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| VIDEO | VIDEO | — |