Kling Image(First Frame) to Video

The most direct Kling node there is: one starting image, a prompt, and Kling animates that image forward into a video. No end frame, no motion reference, no storyboards - this is the plain image-to-video workhorse, the node that represents "Kling" to most people. It's also the one whose defaults have aged gracefully: in current ComfyUI core it runs on the kling-v2-5-turbo model in pro mode, which is the speed-and-quality sweet spot Kling settled on for this kind of straightforward generation.

Like every node in the built-in partner/video category, it's an API node: ships with ComfyUI core, uploads your start frame to Comfy Org's backend, calls Kling's image2video endpoint, and bills your Comfy account in credits. Nothing renders locally. Your image goes up, Kling queues and renders, and the finished clip comes back as a VIDEO.

The inputs that matter

• start_frame - the reference image that becomes the first frame. This is the whole ballgame: Kling's I2V holds the subject's identity from this frame, so the better the image, the better the clip.
• prompt and negative_prompt - both supported, which is a genuine plus in a closed-API world where negatives often get ignored.
• model_name - currently kling-v2-5-turbo. The turbo suffix is doing real work: faster and cheaper than the old v2-master defaults.
• mode - pro. There's a single option here in current builds; it's the quality mode, and it's what you want.
• cfg_scale - 0 to 1, default 0.8. Higher = stricter adherence to the prompt; lower = more freedom. For I2V, a high-ish value keeps the model honest about your instructions.
• aspect_ratio - 16:9, 9:16, 1:1 and friends. duration - a combo pick (5 or 10 seconds).

Outputs: VIDEO, plus a video_id and duration string. The video_id is the useful part - Kling's extend node accepts it, so you can take this clip and continue it past its natural length as part of a longer sequence.

Where people get burned

• The frame is the contract. Kling I2V starts from your image and largely respects it, but it will also move it - camera pushes, subject motion, added detail. If the output wanders from your frame, that's the model doing its job; steer with the prompt rather than fighting the anchor.
• Turbo is a trade, not a free lunch. It's faster and cheaper, and on high-motion or detail-heavy shots the speed shows. For still-life and slow scenes it's nearly indistinguishable from the pricier tier.
• 10 seconds is the ceiling here. This node's duration cap is a combo at 10s; the Kling 3.0 Omni nodes stretch to 15. If you need longer single takes, you're in the wrong node - or you should chain via video_id extension.
• Paid per generation. The badge scales with duration. Draft the prompt once, render once.

For the "I have a picture and I want it to move" request, this is the node - simple, turbo-priced, and it feeds the extension chain via its video_id. It's the right first Kling node to learn on.