Synthesia API

Generates professional presenter videos by synthesizing realistic AI avatar performances synchronized to input text or audio scripts. The system processes text input through a speech synthesis pipeline, generates corresponding facial animations and lip movements, and composites the avatar into a video output with configurable scene duration (up to 5 minutes per scene, 150 scenes max per project). Supports 140+ languages with automatic language detection and voice selection.

Converts PowerPoint presentations (.pptx format) into editable video projects by parsing slides, extracting text and images, and automatically generating scenes with speaker notes as scripts. The system supports files up to 1GB with maximum 150 slides, converting each slide into an editable scene with text, images, videos, and shapes preserved as individual elements. Animations and transitions are not imported; tables are rendered as static non-editable elements.

Generates video scene structures and scripts from unstructured input (documents, URLs, or prompts) using an AI assistant that parses content, segments it by paragraph breaks, and creates a structured scene outline with suggested scripts. Supports document upload (.ppt, .pptx, .pdf, .doc, .docx, .txt up to 50MB), URL content extraction (up to 4,500 words), or direct prompt input. The system automatically segments content into scenes and generates speaker scripts for each scene.

Provides pre-built video templates with standardized layouts, color schemes, fonts, and branding elements that can be applied across multiple videos for visual consistency. Templates define scene structure, background styling, avatar placement, and text formatting rules. Users can select a template when creating a video, and all scenes inherit the template's styling automatically.

Enables creation of custom AI avatars beyond the default library, allowing organizations to use branded or personalized presenter appearances. The custom avatar creation process is not fully documented, but the system supports storing, versioning, and selecting custom avatars for use in video generation. Custom avatars can be applied to any video project and are managed through an avatar library interface.

Generates videos in 140+ languages with automatic language detection from input text and corresponding voice/avatar selection. The system maps input language to available voice models and avatar configurations, synthesizing speech in the detected language with lip-sync animation. Supports language-specific text processing (punctuation, phonetics) for accurate speech synthesis.

Manages video generation as an asynchronous workflow where projects are created, configured, and submitted for processing, with state tracking throughout the generation pipeline. The system stores project state (scenes, avatars, scripts, templates) and processes videos in the background, returning project IDs for status polling or webhook callbacks. Supports up to 150 scenes per project with maximum 4 hours total duration.

Enables granular control over individual video scenes, allowing composition of text overlays, background images, embedded videos, and avatar placement within each scene. Scenes support maximum 5 minutes duration and can include multiple elements (text, images, videos, shapes) positioned and styled independently. Text elements support formatting (font, size, color) and can be edited post-import.

Generates or replaces audio tracks in existing videos with AI-synthesized speech in multiple languages. The Dubbing API accepts video input and text scripts, synthesizes speech in specified language, and produces a dubbed video with synchronized audio. Supports 140+ languages and enables rapid localization of existing video content without re-recording.

Manages a centralized library of media assets (images, videos, audio files) that can be reused across multiple video projects. The Assets API enables uploading, organizing, tagging, and retrieving media assets for use in scene composition. Assets are stored in a project-scoped or organization-scoped library and can be referenced by ID in video projects.