Nodes/ComfyUI/Kling 2.6 Text to Video with Audio
ComfyUI Node Runs on cloud

Kling 2.6 Text to Video with Audio

Describe a scene, get sound with it

By Comfy-Org·Created 4 years ago·Updated 32 minutes ago· 129,926
Kling 2.6 Text to Video with Audio
    • VIDEO
    model_name
    prompt
    mode
    aspect_ratio
    duration
    generate_audiotrue

    The Kling Text to Video with Audio node (Kling 2.6) is the text-only sibling of the 2.6 image-to-video-with-audio node, and it's the one that demonstrates Kling's audio trick most purely: type a scene, get back a clip with matching sound already in it. Rain, footsteps, a door creaking - the model generates the audio to go with what it renders. That's a real capability gap versus most video models, and it's why this node (and the whole "with audio" line) exists.

    The pipeline is the standard partner-node shape. Your prompt goes to Kling's servers through your Comfy account, generation runs in the cloud, and the finished VIDEO returns when the task completes. No local model, no downloads - billed per generation to Comfy account credits, with audio roughly doubling the cost of a silent clip, and the node's price badge reflects that live.

    The inputs that matter:

    • prompt - the positive text prompt. Up to 2,500 characters, and since you're generating audio too, describing the soundscape helps: "ocean waves crashing, seagulls in the distance" gets you a more deliberate audio bed than "beach."
    • model_name - fixed to kling-v2-6.
    • mode - pro is the only option.
    • aspect_ratio - 16:9, 9:16, or 1:1.
    • duration - 5 or 10 seconds.
    • generate_audio - the toggle that matters, on by default. Flip it off for silent footage at a lower price.

    Single VIDEO output into Save Video or Preview Video.

    The practical knobs are few, so the decisions are: how long, what ratio, and whether to pay for sound. The gotcha is cost discipline - with the audio toggle defaulting to on, a batch of re-rolls quietly bills as if every attempt were premium. If you're iterating on motion first, kill generate_audio until you're happy, then turn it on for the keeper. Also keep the model line in perspective: 2.6 sits below Kling 3.0, which generates native audio on its turbo model and can't even turn it off. If you're choosing between the two, the 3.0 path (KlingVideoNode) is where the newer audio-video work is happening.

    Standard partner-node housekeeping: be signed in to your Comfy account or the node fails immediately, and if the API-node family litters your search, --disable-api-nodes removes them all.

    Categorypartner/video/Kling

    Inputs (6)

    NameTypeDefaultDescription
    model_nameCOMBO1 options: kling-v2-6
    promptSTRINGPositive text prompt.
    modeCOMBO1 options: pro
    aspect_ratioCOMBO3 options: 16:9, 9:16, 1:1
    durationCOMBO2 options: 5, 10
    generate_audioBOOLEANtrue

    Outputs (1)

    NameTypeDescription
    VIDEOVIDEO