Video Generation

Generate a video from a text prompt, with optional reference images, audio, text. Results land on the canvas at the click position.

Set up

Add API Key first

• WaveSpeed – Best, all models, recommended (Get API Key)
• GeminiGen — low-cost, but can be slower (good for Google Veo, Seedance)
• Kling – if you want to use direct Kling API program

Open the panel

Click Generate Video in the bottom toolbar.

Text-to-video

Leave the Input Image slot empty. Write a prompt and click Generate.

Image-to-video

Attach a reference image in the Input Image slot. Two modes:

• First-frame only — Kling animates the reference as the opening frame.
• First and last frame — attach a second reference as the end frame; Kling interpolates between them.

Image-to-video is the most reliable way to keep characters and scenes consistent across clips.

Parameters

In Advanced Settings:

• Duration — 5 s or 10 s, model-dependent.
• Aspect ratio — 16:9, 9:16, 1:1, dependent on model.
• CFG scale — how strictly the model follows the prompt; higher = stricter.
• Negative prompt — text describing what to avoid.

Cost

App will show estimated price. When you select model, it will send request to actual price and can update estimated price.

Omni Models

Currently, Seedance can be used in Omni mode. It means that you can reference images, audio, and video all together and ask to build a video based on them

Upscale models

Generation in full size can be high cost, so most of the time you can generate in low resolution and then use an Upscale model on such video. Total cost will be 2-3 times less than generating in full resolution from start.

Recommendations

Even if you generate in low resolution, it’s recommended to use Images as sources in high 2k or 4k resolution.

LipSync Models

If your goal is to make lipsync, use Avatar models. We recommend SkyReal from Wavespeed (Get API Key). Another way to make Lypsinc is using Omni Models (quality can be higher but price will be 3 times more)

Where to go next

• WaveSpeed — alternate provider for video models.
• Video Editor Overview — stitch generated clips into a timeline.
• AI Fill — generate a bridging clip between two frames.