Introduction

Pixamake removes captions, hooks, titles and other text added in editing from video ads, frame by frame. Every frame that carries text is rebuilt with AI, then checked: leftover text, halos and blur are refused before a frame is kept. Text that is part of the scene, like packaging, labels or a phone screen, stays in the video.

You can use Pixamake from the web app, from the REST API, or from an AI agent through the MCP server.

How a video is processed

1. Input. You upload an MP4 or MOV file, or give a public URL.
2. Reading. AI reads the captions shown on screen and the spoken words. This list tells the engine exactly which text to erase. You can also provide it yourself with texts.
3. Rebuilding. Each frame carrying text is regenerated on GPUs, then verified.
4. Delivery. You get an MP4 at the original resolution (up to 1080p), with the original audio.

Core concepts

Concept	Description
Workspace	Holds your plan, credits, videos, API keys and webhook. Teams share one workspace on the Scale plan.
Upload	A slot to send a video file directly to storage, then attach it to a job.
Job	One video being cleaned. It moves from queued to succeeded, failed or canceled.
Credits	150 credits per minute of video, billed per started second. Reserved when a job starts, returned if it fails.

API at a glance

Method	Endpoint	Purpose
GET	/v1/account	Plan, credits and limits
GET	/v1/estimate	Credits for a given duration
POST	/v1/uploads	Get a presigned URL to upload a file
POST	/v1/jobs	Start cleaning a video
GET	/v1/jobs	List jobs
GET	/v1/jobs/{id}	Status, progress and result
POST	/v1/jobs/{id}/cancel	Stop a job and get the credits back

The base URL is https://pixamake.ai. All requests and responses use JSON. The full reference is available as an interactive page and as an OpenAPI 3.1 file.

Limits

• MP4 or MOV, up to 15 minutes, 2 GB and 1080p (1920 by 1080 in either orientation).
• API access, webhooks and MCP are included in the Growth and Scale plans.