Kling 3.0 Turbo (BudgetPixel)
The fast Kling that ships its own audio
- image
- video
Kuaishou's Kling is the video model that the community kept pointing at every time it asked "why can't open weights do this yet?" - it's been the paid-service reference point for years. Kling 3.0 Turbo is the speed tier: fast text-to-video and image-to-video (start frame only) in 3–15 seconds, at 720p or 1080p, with one notable built-in: native audio is always included. No toggle, no "maybe" - every clip comes with sound, which is both a feature and a workflow detail you need to know about before it surprises you.
At 120 credits per second it's priced below the Pro and 4K tiers, which is the point: Turbo is the "I need to see this idea move" node. When you're iterating on a video concept, spending Standard-tier money per draft is how you go broke fast; Turbo is the compromise that keeps iteration affordable without dropping into the bottom shelf.
How it works
Standard BudgetPixel video flow: POST your prompt (and optional start-frame image) into the Kling pipeline, the node polls the status endpoint until succeeded, and the finished MP4 downloads back as a core VIDEO. The start-frame-only constraint is worth internalizing - Turbo doesn't take an end frame, so you can't pin down where the clip finishes. You get a beginning and let the model run. The included audio also means the returned video is a complete file, not a silent render awaiting a soundtrack.
The inputs that matter
- prompt - required, multiline. Describe the motion and the scene.
- image - optional start frame for image-to-video. With it, the model animates from that frame; without, it's pure text-to-video.
- length_seconds - any duration from 3 to 15. This is the price lever: per-second billing means a 15-second run is five times a 3-second run.
- resolution - 720p or 1080p; priced per tier.
- aspect_ratio - framing for the output.
Output is a single VIDEO - Save Video or frame-extract it. Use a recent ComfyUI (2025+) for the video type.
Install
Manager (search "BudgetPixel") or:
cd ComfyUI/custom_nodes
git clone https://github.com/budgetpixel-ai/comfyui-budgetpixel.git
pip install -r comfyui-budgetpixel/requirements.txt
Restart, set BUDGETPIXEL_API_KEY or budgetpixel.json ({"api_key": "bpx_live_..."}) in the pack folder. Video inputs upload via POST /v1/uploads; requests is the only dependency, no model files.
Where people get burned
The always-on audio is the gotcha that bites the most. If your downstream pipeline expects silent clips (a video-to-music node, a dialogue pass, a silent loop), you'll need a mute step - there's no parameter to turn the sound off. Second, per-second billing plus a 15-second cap is the same math as the rest of the video shelf: check Cost Estimate with length_seconds before queueing long renders. Third, Turbo is a speed tier, not a quality tier - if you need the flagship look, the Standard, Pro, or 4K variants of Kling 3.0 are a call away. And remember force_rerun when you want a genuine re-roll instead of a cached clip.
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | Text description of the video. | |
| aspect_ratioopt | COMBO | 16:9 | Aspect ratio. Ignored when a start image already sets the frame. |
| imageopt | IMAGE | Optional start frame for image-to-video. Provide a public image URL, a data URI, raw base64, or an uploaded-file URL from POST /v1/uploads. Omit for text-to-video. | |
| length_secondsopt | INT | 53–15 | Video length in seconds (3–15). |
| resolutionopt | COMBO | 720p | Output resolution. Pricing varies by resolution — see `resolution_pricing` in GET /v1/models. |
| force_rerunopt | BOOLEAN | false | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| video | VIDEO | — |