Elai

Converts raw text input or topic prompts into full video scripts using GPT-based language models, then automatically generates storyboards and renders presenter-led video with synchronized avatar animation and voiceover. The system chains text generation → slide/scene extraction → avatar animation synthesis → audio-visual synchronization in a single browser-based workflow.

Accepts a web URL as input, automatically extracts text content from the page, generates a video script from that content, and renders a complete presenter-led video. The extraction mechanism (likely DOM parsing or content API) feeds into the text-to-video pipeline, enabling one-click conversion of blog posts, articles, or web pages into video format.

Renders videos in 4K Ultra HD resolution (3840x2160) on Team tier and above, while Free and Creator tiers are limited to 1080p Full HD (1920x1080). The rendering pipeline supports both resolutions with automatic quality optimization (bitrate, codec, compression) based on tier. Higher resolution output is available for premium subscribers seeking broadcast-quality or high-fidelity video.

Provides Enterprise tier users with Brand Kit functionality (custom fonts, colors, logos) and Workspace management for multi-team organization. Brand Kit enables consistent visual styling across all videos created by an organization, while Workspaces allow separate teams or departments to manage their own video libraries and settings within a single enterprise account. These features are integrated into the rendering pipeline and user management system.

Provides Enterprise tier users with SSO integration (likely SAML 2.0 or OAuth 2.0) for centralized identity management and authentication. Users log in via their organization's identity provider (Okta, Azure AD, Google Workspace, etc.) rather than creating separate Elai credentials. SSO integration is managed at the account level and applies to all team members within an enterprise workspace.

Provides access to premium voices (beyond the standard 450+ voices) on Team tier and above. Premium voices offer higher quality, more natural-sounding synthesis, and may include celebrity or branded voices. Voice customization options (if available) may include speech rate, tone, or emphasis adjustments, though the extent of customization is unknown.

Accepts presentation files (format unspecified, likely PowerPoint or Google Slides) as input and automatically converts slides into a video with synchronized avatar narration. The system likely parses slide content, extracts text/speaker notes, generates or uses existing voiceover, and animates avatar transitions between slides to create a presenter-led video.

Synthesizes natural-sounding voiceover in 75+ languages using a voice synthesis engine (likely neural TTS) with access to 450+ pre-built voices. Additionally supports voice cloning, where users record a short audio sample (30-60 seconds typical) and the system generates synthetic speech in that user's voice for personalized narration. Voice selection and cloning are integrated into the video rendering pipeline.

Provides 80+ pre-built avatars (diverse in appearance, gender, age, ethnicity) for immediate use, plus four types of custom avatars: Selfie (user video), Studio (professional video), Photo (static image-based), and Animated mascot. Custom avatars are created by uploading video or image assets; the system then animates these avatars using the same synthesis engine as pre-built avatars, enabling lip-sync and gesture animation synchronized to voiceover.

Automatically converts video scripts into visual storyboards and slides without manual input. The system parses script text, extracts key scenes or topics, generates corresponding slide layouts, and sequences them for video rendering. The algorithm (proprietary, undisclosed) determines slide timing, visual hierarchy, and transitions based on script content.

Enables creation of multiple video variations at scale by inserting custom variables (names, company names, personalized messages, graphics) into a template video. The system uses an API-based workflow where users define variables, provide a data source (CSV, API, or manual list), and the platform renders individual videos for each data row. Variables are substituted into scripts, slides, and graphics, enabling personalized outreach campaigns without manual per-video editing.

Provides a browser-based text editor with GPT-based suggestions and refinement capabilities for script writing and editing. Users can write raw scripts, request AI-generated suggestions for improvements (tone, clarity, length), and iterate on scripts before rendering. The editor integrates with the text-to-video pipeline, allowing users to refine scripts before storyboarding and rendering.

Generates images from text prompts using an AI image generation model (likely DALL-E or similar) and inserts them into video slides. Additionally integrates a stock media library (videos, images, Lottie animations) that users can search and insert into videos. Generated images and stock media are automatically sized and positioned to fit slide layouts during rendering.

Renders videos in three aspect ratios: 16:9 (landscape/widescreen), 9:16 (vertical/mobile), and 1:1 (square/social media). The system automatically adapts slide layouts, avatar positioning, and text sizing for each aspect ratio without requiring separate video creation. Users select aspect ratio at render time, and the platform handles layout reflow and optimization.