Molmo2 Video Understanding API Documentation

Molmo2 Video Understanding API Documentation

Playground

Molmo2-4B Video Understanding: Analyze videos with specialized tasks (general, summary, analysis, counting, scene description). Open-source vision-language model with temporal understanding. Ready-to-use REST API, no cold starts, duration-based pricing.

Features

Analyze and understand video content with Molmo2 Video Understanding. This intelligent video analysis model performs various tasks including summarization, scene description, object counting, and detailed analysis — perfect for video cataloging, content moderation, and automated video workflows.

Why It Works Great

• Multiple task modes: Summary, analysis, counting, scene description, and general Q&A.
• Custom instructions: Add specific focus areas or questions.
• Extended video support: Analyze videos up to 2 minutes long.
• Structured output: Get organized, task-specific results.
• Affordable: Starting at just $0.005 per video.
• Versatile analysis: From quick summaries to detailed breakdowns.

Parameters

How to Use

1. Upload your video — drag and drop or paste a public URL.
2. Select task — choose the type of analysis you need.
3. Add instructions (optional) — specify focus areas or custom questions.
4. Run — click the button to analyze.
5. Review results — get structured analysis output.

Pricing

Per 5-second billing with minimum charge for videos ≤5 seconds. Maximum billable duration is 120 seconds.

Task Modes

Best Use Cases

• Video Cataloging — Automatically generate descriptions for video libraries.
• Content Moderation — Analyze video content for review workflows.
• Accessibility — Create text descriptions for visually impaired users.
• Search & Discovery — Generate metadata for video search systems.
• Analytics — Count objects, people, or events in footage.
• Summarization — Create quick summaries for long-form content.

Example Instructions

• “Focus on the people in the video and describe their actions.”
• “Count how many cars appear in this video.”
• “Summarize the main events in chronological order.”
• “Describe the setting and atmosphere of each scene.”
• “Identify any text or logos visible in the video.”
• “What products are being demonstrated?”

Pro Tips for Best Results

• Choose the appropriate task mode for your specific need.
• Use “text” parameter to focus analysis on specific elements.
• Combine “general” task with custom questions for flexible Q&A.
• Use “counting” for metrics like people, objects, or occurrences.
• “scene_description” works great for accessibility and content tagging.
• Keep videos under 2 minutes for optimal processing.

Notes

• Maximum supported video duration is 120 seconds (2 minutes).
• If using a URL, ensure it is publicly accessible.
• Processing time scales with video length.
• Different tasks produce different output formats optimized for their purpose.

Duration limit

The maximum supported video duration is 2 minutes.

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'
{
  "video": "https://interactive-examples.mdn.mozilla.net/media/cc0-videos/flower.mp4",
  "task": "general"
}
JSON
)

# 1. Submit the prediction.
SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/molmo2/video-understanding" \
  -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