Magic Video

Magic is a video generation model designed to transform text prompts, static images, or existing videos into high-quality, temporally consistent video content. The model is built for creators, product teams, marketers, and developers who need controllable, scalable, and production-ready AI video generation.

Technical Specifications

• Model Type: Diffusion-based multimodal video transformer
• Input Modalities: Text (natural language prompts); Images (single or multiple frames); Video clips

• Output: Video sequence (latent-to-pixel decoded)
• Temporal Modeling: Multi-head temporal attention
• Resolution Support: 720p; 1080p

• Frame Rate: 24–30 FPS
• Clip Length: Short to mid-range video generation

Text-to-Video API

This model converts descriptive language into original video sequences.

• Core Technology: Advanced diffusion models and transformer architectures that interpret nuanced language to generate coherent visual narratives.
• Input: Text prompts (English). Supports complex, descriptive prompts detailing subject, action, style, and camera motion.
• Output:
  • Resolution: Up to 1080p (1920x1080).
  • Duration: Configurable clips, typically between 5 to 10 seconds per generation.
  • Frame Rate: 24-30 fps for cinematic smoothness.

Image-to-Video API

This model animates static images by adding realistic, user-directed motion.

• Core Technology: AI that analyzes image composition and depth to apply plausible motion to elements within the frame.
• Input: A single still image (JPG, PNG) and an optional text prompt describing the desired motion.
• Output:
  • Resolution: Matches input up to 1080p; requires source images of at least 720p for optimal quality.
  • Duration: Standard outputs of 5 or 10 seconds.
  • Motion Types: Can simulate camera pans, zooms, parallax effects, and natural element movement (e.g., swaying leaves, flowing water)

Video-to-Video API

This model applies stylistic transformations or targeted edits to existing video footage.

• Core Technology: Temporal-aware diffusion models that maintain consistency across video frames during editing.
• Input: A source video clip and a text prompt defining the edit (e.g., "apply cyberpunk style," "change background to a forest").
• Output:
  • Resolution: Maintains input video resolution.
  • Key Features: Style transfer, background replacement, object addition/removal, and resolution upscaling.

API Pricing

• $0.65 per generation

Key Features

• High-Resolution Output: Generate professional-grade videos up to 1080p, suitable for social media and marketing materials.
• Native Audio Generation: Create videos complete with sound effects and ambient audio for a fully immersive result.
• Intuitive Motion Control: Guide animations through descriptive language, making sophisticated motion accessible.

‍

Use Cases

Creative & Media Production

Magic enables creators to quickly turn ideas into visually rich video content without traditional animation or filming workflows. It is well suited for short films, music visuals, experimental art, and cinematic concept exploration with full creative control.

Marketing & Advertising Content

Marketing teams use Magic to generate promotional videos, product teasers, and branded visuals at scale. The model allows fast iteration of concepts, styles, and messaging without costly reshoots or complex post-production.

Social Media & Short-Form Video

The model is optimized for generating eye-catching short videos tailored for social platforms. It supports rapid content production, trend-based visuals, and consistent brand aesthetics across multiple formats.

Video Transformation & Re-Styling

With Video-to-Video, Magic enables creators to reimagine existing footage in new styles or contexts. This is useful for localization, brand adaptation, visual upgrades, and creative reinterpretation of existing assets.