0.2.0

Video Generation

Generate videos from the Decart Android SDK. Submit a video, image, or text prompt - get back a transformed video.

Quick Start — Video-to-Video with lucy-2-v2v

val client = DecartClient(
    context = context,
    config = DecartClientConfig(apiKey = "your-api-key")
)

val input = VideoEditInput(
    prompt = "Transform into Studio Ghibli anime style",
    data = FileInput.fromUri(videoUri),      // from a file picker
    enhancePrompt = true,
)

// Option 1: Simple poll-and-wait
val result = client.queue.submitAndPoll(VideoModels.LUCY_2_V2V, input)
if (result is QueueJobResult.Completed) {
    File("output.mp4").writeBytes(result.data)
}

// Option 2: Reactive Flow with progress updates
client.queue.submitAndObserve(VideoModels.LUCY_2_V2V, input).collect { update ->
    when (update) {
        is QueueJobResult.InProgress -> Log.d("TAG", "Status: ${update.status}")
        is QueueJobResult.Completed -> File("output.mp4").writeBytes(update.data)
        is QueueJobResult.Failed -> Log.e("TAG", "Failed: ${update.error}")
    }
}

client.release()

Supported Models

Model	Type	Input
lucy-2-v2v	Video Edit	Video + prompt + optional reference image
lucy-pro-v2v	Video Edit (Pro)	Video + prompt + optional reference image
lucy-fast-v2v	Video Edit (Fast)	Video + prompt + optional reference image
lucy-restyle-v2v	Video Restyle	Video + prompt or reference image
lucy-pro-t2v	Text to Video	Prompt only (landscape/portrait)
lucy-pro-i2v	Image to Video (Pro)	Image + prompt
lucy-dev-i2v	Image to Video (Dev)	Image + prompt
lucy-motion	Motion Video	Image + trajectory path

Each model has a dedicated input type (VideoEditInput, VideoRestyleInput, TextToVideoInput, ImageToVideoInput, MotionVideoInput) with built-in validation.

What's New

• Queue API — submit, status, result, submitAndPoll, submitAndObserve (Kotlin Flow)
• Streaming uploads — large videos upload without loading into memory
• Model-adaptive sample app — "Video" tab auto-adjusts UI fields per model

What's Changed

• Add batch video queue support and typed inputs across video models by @AdirAmsalem in #1

Full Changelog: 0.1.0...0.2.0