Moondream3 Preview Point

Moondream3 Preview Point

Playground

Moondream3 Point finds objects in images and returns precise coordinate points for computer vision tasks, enabling accurate point localization. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

Moondream 3 Point is a vision-language model designed to identify and describe specific objects within an image using natural language. Instead of returning coordinates, it provides a concise textual description of the detected object, making it ideal for lightweight interactive queries and content understanding.

✨ Key Features

• Locate and Describe Objects Enter a short text query (e.g., “hat”, “watch”, “phone”) and receive a natural-language description of that item in context.

• Fast Single-Object Queries Optimized for fast, low-latency inference — perfect for real-time applications.

• Readable Natural Output The model outputs a fluent English sentence describing the object’s appearance, position, and context.

• Multilingual Understanding Capable of recognizing and describing objects in a wide range of visual scenarios.

⚙️ Example Usage

Locate & Describe “Hat”

{
 "image": "https://example.com/photo.jpg",
 "prompt": "hat"
}

Example Response

{
 "answer": "The woman is wearing a pink baseball cap with a strap across her forehead. She is also wearing large silver hoop earrings and a pink fuzzy sweater."
}

💡 Best Practices

• Use concise object names (e.g., “hat”, “car”, “tree”) for more accurate detection.

• For precise bounding boxes or coordinates, use:

• Moondream 3 Detect — returns x_min, y_min, x_max, y_max bounding boxes.

• A coordinate-enabled version of Moondream 3 Point (coming soon).

• Supported formats: JPEG, PNG, WebP

• Maximum image size: 10 MB

💰 Pricing

• $0.001 per request
• Volume and enterprise pricing available upon request.

📝 Notes

• The current endpoint returns descriptive text in JSON format:

{"answer": "..."}

— it does not output coordinates.

• For small or occluded objects, use higher-resolution input or switch to the Detect model for better spatial precision.

Authentication

For authentication details, please refer to the Authentication Guide.

API Endpoints

Submit Task & Query Result

set -euo pipefail

export WAVESPEED_API_KEY="your-api-key"

REQUEST_BODY=$(cat <<'JSON'
{
  "prompt": "A cinematic ocean wave at sunrise, highly detailed",
  "image": "https://interactive-examples.mdn.mozilla.net/media/cc0-images/painted-hand-298-332.jpg"
}
JSON
)

# 1. Submit the prediction.
SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/moondream3-preview/point" \
  -H "Authorization: Bearer ${WAVESPEED_API_KEY}" \
  -H "Content-Type: application/json" \
  -d "${REQUEST_BODY}")

TASK=$(printf '%s' "${SUBMIT_RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
PREDICTION_ID=$(printf '%s' "${TASK}" | jq -r '.id // empty')
if [ -z "${PREDICTION_ID}" ]; then
  printf 'Submission response did not contain a prediction id
' >&2
  exit 1
fi
RESULT_URL=$(printf '%s' "${TASK}" | jq -r '.urls.get // empty')
if [ -z "${RESULT_URL}" ]; then RESULT_URL="https://api.wavespeed.ai/api/v3/predictions/${PREDICTION_ID}/result"; fi

# 2. Poll until the prediction finishes.
while true; do
  RESPONSE=$(curl --silent --show-error --fail-with-body \
    "${RESULT_URL}" \
    -H "Authorization: Bearer ${WAVESPEED_API_KEY}")
  RESULT=$(printf '%s' "${RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
  STATUS=$(printf '%s' "${RESULT}" | jq -r '.status // empty')

  case "${STATUS}" in
    completed) printf '%s\n' "${RESULT}" | jq '.outputs'; break ;;
    failed|cancelled|timeout) printf '%s\n' "${RESULT}" | jq . >&2; exit 1 ;;
    created|processing) sleep 2 ;;
    *) printf 'Unexpected status: %s
' "${STATUS}" >&2; exit 1 ;;
  esac
done

Parameters

Task Submission Parameters

Request Parameters

Response Parameters

Result Request Parameters

Result Response Parameters