Generate creator-style product stills, upscale and polish one, then animate it into a short UGC video.

UGC Product Video

Use this workflow when you want a creator-style product ad. It starts from your trained creator model, produces a recognizable product still, upscales the best still, improves it with a known image model, and uses the final image as the first frame for a short video.

Before you start

Complete Setup and Training first. These snippets assume modelId or model_id, triggerImageGeneration, triggerImageUpscaling, triggerVideoGeneration, waitForJob, and firstImage are already available.

Workflow

Generate a product still with POST /images/generation.
Upscale the selected still with POST /images/upscaling.
Create a polished ad still with POST /images/generation.
Animate the polished still with POST /videos/generation.

Generate the creator product still

This uses the trained model ID, so the generated image should preserve the trained identity while following the product prompt.

const trainedImageJobId = await triggerImageGeneration({
  modelID: modelId,
  prompt:
    'authentic UGC product photo, creator holding a skincare bottle near a bathroom mirror, natural phone camera lighting',
  width: 1024,
  height: 1024,
  numImages: 2,
  enhanceFace: true,
})

const trainedImage = firstImage(await waitForJob(trainedImageJobId), 'UGC trained image')

trained_image_job_id = trigger_image_generation(
    {
        "modelID": model_id,
        "prompt": "authentic UGC product photo, creator holding a skincare bottle near a bathroom mirror, natural phone camera lighting",
        "width": 1024,
        "height": 1024,
        "numImages": 2,
        "enhanceFace": True,
    }
)

trained_image = first_image(wait_for_job(trained_image_job_id), "UGC trained image")

Upscale the selected still

Upscale the best generated still before using it as a reference image for a cleaner campaign version.

const upscaleJobId = await triggerImageUpscaling({
  imageID: trainedImage.id,
})

const upscaledImage = firstImage(await waitForJob(upscaleJobId), 'UGC upscale')

upscale_job_id = trigger_image_upscaling(
    {
        "imageID": trained_image["id"],
    }
)

upscaled_image = first_image(wait_for_job(upscale_job_id), "UGC upscale")

Improve the still with a public image model

Use a public image model such as gpt-image-2 for a more polished ad still. The upscaled image becomes the visual reference.

const polishedImageJobId = await triggerImageGeneration({
  modelID: 'gpt-image-2',
  prompt:
    'polished social ad still, creator holding the skincare bottle, clean bathroom counter, premium but natural UGC style',
  aspectRatio: '16:9',
  numImages: 1,
  referenceImageURLs: [upscaledImage.url],
  options: {
    openai: {
      imageSize: '1K',
      outputFormat: 'jpeg',
      quality: 'high',
    },
  },
})

const polishedImage = firstImage(await waitForJob(polishedImageJobId), 'UGC polished image')

polished_image_job_id = trigger_image_generation(
    {
        "modelID": "gpt-image-2",
        "prompt": "polished social ad still, creator holding the skincare bottle, clean bathroom counter, premium but natural UGC style",
        "aspectRatio": "16:9",
        "numImages": 1,
        "referenceImageURLs": [upscaled_image["url"]],
        "options": {
            "openai": {
                "imageSize": "1K",
                "outputFormat": "jpeg",
                "quality": "high",
            },
        },
    }
)

polished_image = first_image(wait_for_job(polished_image_job_id), "UGC polished image")

Generate the UGC video

Use the polished image as first_frame so the video starts from the campaign still.

const videoJobId = await triggerVideoGeneration({
  modelID: 'seedance-2.0',
  prompt:
    'short UGC product video, creator lifts the skincare bottle, smiles, and points to the label, handheld phone camera energy',
  numVideos: 1,
  referenceImages: [
    {
      url: polishedImage.url,
      role: 'first_frame',
    },
  ],
  options: {
    seedance: {
      duration: 6,
      ratio: '16:9',
      resolution: '1080p',
      generateAudio: false,
    },
  },
})

const videoJob = await waitForJob(videoJobId, 8000)
const video = videoJob.videos?.[0]

video_job_id = trigger_video_generation(
    {
        "modelID": "seedance-2.0",
        "prompt": "short UGC product video, creator lifts the skincare bottle, smiles, and points to the label, handheld phone camera energy",
        "numVideos": 1,
        "referenceImages": [
            {
                "url": polished_image["url"],
                "role": "first_frame",
            },
        ],
        "options": {
            "seedance": {
                "duration": 6,
                "ratio": "16:9",
                "resolution": "1080p",
                "generateAudio": False,
            },
        },
    }
)

video_job = wait_for_job(video_job_id, interval_seconds=8)
video = video_job.get("videos", [])[0]

Production notes

Store the generated image.id values so you can upscale, inspect, or reuse assets later.
Keep the product prompt specific about object placement, hand position, and scene style.
Use referenceImageURLs for known-image-model workflows and referenceImages with role: "first_frame" for image-to-video workflows.