Sora 2

OpenAI's flagship video and audio generation model that creates richly detailed, dynamic video clips with synchronized audio from natural language prompts or images.

Property	Value
Model ID	`openai/sora-2`
Provider	Replicate
Type	Video generation (text-to-video, image-to-video, audio-sync)

Basic Usage

Sora 2 is a versatile video model that works with both text prompts and optional start images:

import { compose, generateVideo, videoModel } from "@synthome/sdk";

const execution = await compose(
  generateVideo({
    model: videoModel("openai/sora-2", "replicate"),
    prompt:
      "A serene beach sunset with waves gently lapping at the shore, seagulls calling in the distance",
  }),
).execute();

Options

Option	Type	Default	Description
`prompt`	`string`	required	Text description of the video scene
`aspectRatio`	`"16:9" \| "9:16"`	-	Video aspect ratio (landscape or portrait)
`duration`	`number` (4-12)	-	Video duration in seconds
`resolution`	`"720p" \| "1080p"`	-	Output video resolution
`image`	`string`	-	Starting frame image URL (image-to-video)

Video Generation

Text-to-Video

Create videos from text descriptions:

generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt:
    "An orange tabby cat knocks over a ceramic mug on a wooden table, with the sound of ceramic breaking, in warm kitchen lighting",
  duration: 8,
  resolution: "1080p",
});

Image-to-Video

Generate videos starting from an image:

generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt: "The tree sways gently in the wind, leaves rustling",
  image: "https://example.com/tree.jpg",
  duration: 6,
});

Audio Synchronization

Sora 2 automatically generates synchronized audio including:

Background sounds (ambient noise, environment audio)
Sound effects (matching on-screen actions)
Dialogue (if specified in prompt)
Music (when appropriate to the scene)

// Includes synchronized audio
generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt:
    "A coffee shop with barista making drinks, espresso machine hissing, gentle background chatter, jazz music playing",
  duration: 10,
});

Aspect Ratios

Choose landscape or portrait formats for different platforms:

// Landscape (16:9) - YouTube, desktop
generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt: "A panoramic mountain vista at sunrise",
  aspectRatio: "16:9",
});

// Portrait (9:16) - TikTok, Instagram Reels, mobile
generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt: "A close-up of blooming flower",
  aspectRatio: "9:16",
});

Note: The SDK automatically converts standard aspect ratios (16:9, 9:16) to Replicate's format (landscape, portrait). You always use the standard notation in your code.

Resolution

Select output quality based on your needs:

// 720p - Faster generation, smaller files
generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt: "Street scene with cars and pedestrians",
  resolution: "720p",
});

// 1080p - Higher quality, larger files
generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt: "Detailed architectural shot with intricate textures",
  resolution: "1080p",
});

Duration

Generate videos from 4 to 12 seconds:

// Short video (4-6 seconds)
generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt: "A quick magic sparkle effect",
  duration: 5,
});

// Longer video (8-12 seconds)
generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt:
    "A complete story arc: character walks, encounters obstacle, overcomes it",
  duration: 12,
});

Cinematic Control

Sora 2 understands cinematic terminology for professional-looking results:

generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt:
    "Cinematic IMAX-scale scene: wide establishing shot of futuristic city, dolly in towards main building, warm morning light, dramatic shadows",
  aspectRatio: "16:9",
  duration: 10,
  resolution: "1080p",
});

Examples

Marketing Content

generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt:
    "Elegant product shot of luxury watch, close-up rotating on velvet surface, soft spotlight, ambient luxury showroom sounds",
  resolution: "1080p",
  duration: 8,
});

generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt:
    "Upbeat cooking tutorial: chef chopping vegetables, sizzling sounds in pan, energetic background music, bright kitchen lighting",
  aspectRatio: "9:16",
  resolution: "720p",
  duration: 6,
});

Educational Content

generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt:
    "Scientific animation: water cycle, clouds forming, rain falling, river flowing, calm narrator voice explaining process",
  duration: 12,
  resolution: "1080p",
});

Creative Animation

generateVideo({
  model: videoModel("openai/sora-2", "replicate"),
  prompt:
    "Studio Ghibli style: young girl on flying bicycle over countryside, whimsical music, birds chirping, gentle wind sounds",
  duration: 10,
});

Best Practices

Be Specific: Include details about lighting, movement, and sounds in your prompt
Audio Descriptions: Mention specific sounds ("espresso machine hissing", "footsteps crunching") for better audio sync
Duration: Shorter videos (4-8 seconds) tend to have better temporal consistency
Physics: Sora 2 has improved physics simulation - describe realistic movements for best results
Multi-Shot: For sequences, clearly delineate each shot: "Shot 1 (0-4s): wide shot. Shot 2 (4-8s): close-up"
Cinematic Terms: Use terms like "dolly in", "pan left", "handheld camera" for precise control

Limitations

Generation time can be several minutes depending on complexity
Temporal consistency improves with shorter clips
Very detailed text rendering may have artifacts
Some highly complex scenarios may not render exactly as described
Requires organization verification for OpenAI API keys

Audio Capabilities

Sora 2 generates sophisticated audio alongside video:

Ambient sounds: Environment noises, weather, background activity
Sound effects: Actions, impacts, movements synchronized with video
Music: Background music matching the scene mood and style
Dialogue: Character speech with lip synchronization
** Foley**: Detailed sound effects for enhanced realism

The audio is automatically embedded in the MP4 output with proper synchronization.

Sora 2

On this page