Video Generation

Generate videos from text, images, or audio — including talking head and lipsync.

Create short video clips from text prompts, animate still images, or produce talking-head videos with lipsync. All results are saved to your Pixa workspace.

Text-to-video

Use the generate_media tool with media_type: "video" and a video model:

Generate a 5-second product reveal video of sneakers rotating on a white background.

Image-to-video

Pass a start frame as an attachment and describe the motion you want:

Animate this product photo with a slow zoom in and subtle background motion.

The model uses the attached image as the opening frame and animates from there.

Talking video

Attach a face image and provide a script — the model animates the face to speak the script with natural lip movement and expression:

Make this face say: "Hi everyone, welcome to our product launch. Today we are introducing something special."

Lipsync

Attach both a face image and an audio file to sync lip movement to existing audio:

Lipsync this face image to the attached voiceover audio.

For lipsync, attach the face image and audio file together as separate attachments.

Video parameters

The generate_media tool accepts:

Parameter	Description	Example
prompt	Description of the video (required)	"slow pan across a sunlit kitchen"
model	Video model ID (required)	kling-v2.6-pro
media_type	Set to "video" for video generation	video
aspect_ratio	Output aspect ratio	16:9, 9:16, 1:1
output_format	File format	mp4, webm
num_variations	Number of variations	1-4
attachments	Start frame, face image, or audio	—

Batch video generation

Generate multiple variations:

Generate 3 variations of a product reveal animation for this image.

To produce the same video in different aspect ratios for different platforms, ask for separate runs:

Generate this product demo in 16:9 for YouTube and 9:16 for TikTok.

Tips

For talking videos, write the script as natural speech. Avoid stage directions or formatting — just write what the person should say.

• Start frame matters. For image-to-video, the quality and composition of your start frame directly affects the output.
• Keep prompts focused. Describe one clear motion or scene rather than a complex sequence.
• Async jobs. Video generation can take several minutes. Use get_job_status with sync: true to wait up to 25 seconds, or poll periodically for longer jobs.

Related

• Video Generation (User Guide) — in-app video workflow
• MCP Server — full MCP tool reference