Seedream 5.0 Pro đã ra mắt | Thử trong Trình tạo ảnh →

Nemotron 3 Nano Omni Video

nvidia /

NVIDIA Nemotron 3 Nano Omni is an open, efficient reasoning model for enterprise agentic workflows, built on a 30B A3B hybrid Transformer-Mamba MoE architecture. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

video-to-text
Đầu vào

Chờ

{
  "output": "A woman with wavy brown hair, wearing a white sweater and blue denim with a belt, is holding a black cassette in her right hand and looking at it. She is in a room with two cream-colored couches on either side, a brown table in the middle, and a lamp on it, along with two photo frames. Behind her, there are three windows with white curtains and brown drapes on the sides."
}

$0.006cho mỗi lần chạy·~166 / $1

Ví dụXem tất cả

Describe the video.

Mô hình liên quan

README

NVIDIA Nemotron-3 Nano Omni Video

NVIDIA Nemotron-3 Nano Omni Video is a multimodal video-language model for understanding and analyzing video content. Provide a video URL and an English prompt, and the model generates a text response for tasks such as video description, scene understanding, event summarization, and visual question answering over time-based media.

Why Choose This?

  • Video understanding with natural-language prompts Ask questions about a video or request summaries, descriptions, and structured analysis in plain English.

  • Temporal scene analysis Understand actions, events, transitions, and visual context across time instead of from a single frame only.

  • Flexible response control Adjust max_tokens, temperature, and top_p to balance response length, determinism, and creativity.

  • Optional system steering Use system_prompt to guide output style, response format, or task behavior for more controlled results.

  • Reasoning mode options Choose between no_think and think depending on your preferred response mode and workflow.

  • Production-ready API Suitable for video analysis pipelines, multimodal assistants, content review systems, and automated media understanding workflows.

Parameters

ParameterRequiredDescription
promptYesEnglish text prompt sent to the model.
video_urlYesURL of the video to analyze.
system_promptNoOptional system prompt used to steer behavior, tone, or response style.
reasoning_modeNoReasoning mode: no_think (default) or think.
max_tokensNoMaximum number of tokens to generate. Default: 1024.
temperatureNoSampling temperature. Lower values are more deterministic. Default: 0.7.
top_pNoNucleus sampling probability mass. Default: 0.95.

How to Use

  1. Provide your video URL — upload or link the video you want the model to analyze.
  2. Write your prompt — ask the model to describe, summarize, explain, compare, classify, or answer questions about the video.
  3. Add a system prompt (optional) — guide the output style, response structure, or task framing.
  4. Choose reasoning mode (optional) — use no_think or think depending on your workflow.
  5. Set generation controls (optional) — adjust max_tokens, temperature, and top_p.
  6. Submit — run the model and review the generated response.

Example Prompt

Describe this video in detail, including the setting, key actions, important scene changes, visible subjects, and the overall mood.

Pricing

Billed by configured max_tokens.

Max TokensCost
1000$0.006
1024$0.0061
2000$0.012
4000$0.024
8000$0.048

Billing Rules

  • Pricing is based on the configured max_tokens value.
  • Cost is $0.006 per 1,000 max tokens.
  • Increasing max_tokens increases cost linearly.
  • prompt, video_url, system_prompt, reasoning_mode, temperature, and top_p do not change pricing directly.

Best Use Cases

  • Video summarization — Generate concise or detailed summaries of video content.
  • Scene and event understanding — Identify key actions, transitions, and important moments over time.
  • Video question answering — Ask targeted questions about what happens in the video.
  • Content review workflows — Inspect uploaded videos for structured analysis, moderation, or categorization tasks.
  • Multimodal assistants — Add video-aware understanding to internal tools, bots, and applications.
  • Research and annotation tasks — Use guided prompts to consistently analyze and label video content.

Pro Tips

  • Write prompts in English for best compatibility.
  • Be specific about the task, such as summarization, scene breakdown, action recognition, or focused question answering.
  • Use system_prompt when you need a consistent output format, such as bullet summaries, labeled sections, or structured JSON-like responses.
  • Keep temperature lower when you want more stable and deterministic answers.
  • Increase max_tokens only when you need longer outputs, since pricing is tied to that value.
  • Ask focused temporal questions like “what happens before and after” or “summarize the sequence of events” for better video-specific results.

Notes

  • Both prompt and video_url are required.
  • prompt must be written in English.
  • Default settings include reasoning_mode = no_think, max_tokens = 1024, temperature = 0.7, and top_p = 0.95.
  • Pricing depends on configured max_tokens, not on other generation settings.

Related Models

Lưu ý:Trang web này sử dụng các mô hình AI do bên thứ ba cung cấp.

Nemotron 3 Nano Omni Video API — Quick start

Grab a WaveSpeedAI API key, then call POST https://api.wavespeed.ai/api/v3/nvidia/nemotron-3-nano-omni/video with your input as JSON. The endpoint returns a prediction id. Start polling the result endpoint around every 2 seconds, increase the interval for long-running tasks, and stop on any terminal status. On completed, read output values from data.outputs. Examples for Nemotron 3 Nano Omni Video below.

HTTP example
set -euo pipefail

: "${WAVESPEED_API_KEY:?Set WAVESPEED_API_KEY}"

REQUEST_BODY=$(cat <<'JSON'
{
    "prompt": "A cinematic shot of a city at sunset, soft golden light",
    "video_url": "https://interactive-examples.mdn.mozilla.net/media/cc0-videos/flower.mp4",
    "reasoning_mode": "no_think",
    "max_tokens": 1024,
    "temperature": 0.7,
    "top_p": 0.95
}
JSON
)

# 1. Submit the prediction.
SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  -X POST "https://api.wavespeed.ai/api/v3/nvidia/nemotron-3-nano-omni/video" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $WAVESPEED_API_KEY" \
  -d "$REQUEST_BODY")

TASK=$(printf '%s' "$SUBMIT_RESPONSE" | jq 'if has("data") then .data else . end')
PREDICTION_ID=$(printf '%s' "$TASK" | jq -r '.id')
if [ -z "$PREDICTION_ID" ] || [ "$PREDICTION_ID" = "null" ]; 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' "$RESPONSE" | jq 'if has("data") then .data else . end')
  STATUS=$(printf '%s' "$RESULT" | jq -r '.status')
  case "$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
Node.js example
const submitUrl = "https://api.wavespeed.ai/api/v3/nvidia/nemotron-3-nano-omni/video";
const apiKey = process.env.WAVESPEED_API_KEY;
if (!apiKey) throw new Error('Set WAVESPEED_API_KEY');

async function requestJson(url, options = {}) {
  const response = await fetch(url, options);
  if (!response.ok) throw new Error(await response.text());
  return response.json();
}

// 1. Submit the prediction.
const body = await requestJson(submitUrl, {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${apiKey}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
        "prompt": "A cinematic shot of a city at sunset, soft golden light",
        "video_url": "https://interactive-examples.mdn.mozilla.net/media/cc0-videos/flower.mp4",
        "reasoning_mode": "no_think",
        "max_tokens": 1024,
        "temperature": 0.7,
        "top_p": 0.95
}),
});
const task = body.data ?? body;
if (!task.id) throw new Error("Submission response did not contain a prediction id");
const resultUrl = task.urls?.get ||
  `https://api.wavespeed.ai/api/v3/predictions/${task.id}/result`;

// 2. Poll until the prediction finishes.
while (true) {
  const resultBody = await requestJson(resultUrl, {
    headers: { "Authorization": `Bearer ${apiKey}` },
  });
  const result = resultBody.data ?? resultBody;
  if (result.status === "completed") {
    console.log(result.outputs);
    break;
  }
  if (["failed", "cancelled", "timeout"].includes(result.status)) throw new Error(JSON.stringify(result));
  if (!["created", "processing"].includes(result.status)) throw new Error("Unexpected status: " + result.status);
  await new Promise(resolve => setTimeout(resolve, 2000));
}
Python example
import json
import os
import time
from urllib.request import Request, urlopen

api_key = os.environ["WAVESPEED_API_KEY"]
headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
payload = {
    "prompt": "A cinematic shot of a city at sunset, soft golden light",
    "video_url": "https://interactive-examples.mdn.mozilla.net/media/cc0-videos/flower.mp4",
    "reasoning_mode": "no_think",
    "max_tokens": 1024,
    "temperature": 0.7,
    "top_p": 0.95
}

def request_json(url, data=None):
    request = Request(url, data=data, headers=headers, method="POST" if data else "GET")
    with urlopen(request) as response:
        return json.load(response)

# 1. Submit the prediction.
body = request_json("https://api.wavespeed.ai/api/v3/nvidia/nemotron-3-nano-omni/video", json.dumps(payload).encode())
task = body.get("data", body)
if not task.get("id"):
    raise RuntimeError("Submission response did not contain a prediction id")
result_url = task.get("urls", {}).get("get") or f"https://api.wavespeed.ai/api/v3/predictions/{task['id']}/result"

# 2. Poll until the prediction finishes.
while True:
    result_body = request_json(result_url)
    result = result_body.get("data", result_body)
    status = result.get("status")
    if status == "completed":
        print(result.get("outputs", []))
        break
    if status in {"failed", "cancelled", "timeout"}:
        raise RuntimeError(result)
    if status not in {"created", "processing"}:
        raise RuntimeError(f"Unexpected status: {status}")
    time.sleep(2)

Nemotron 3 Nano Omni Video API — Frequently asked questions

What is the Nemotron 3 Nano Omni Video API?

Nemotron 3 Nano Omni Video is a NVIDIA model for AI inference, exposed as a REST API on WaveSpeedAI. NVIDIA Nemotron 3 Nano Omni is an open, efficient reasoning model for enterprise agentic workflows, built on a 30B A3B hybrid Transformer-Mamba MoE architecture. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing. You can call it programmatically or try it from the playground above.

How do I call the Nemotron 3 Nano Omni Video API?

POST your input parameters to the model's REST endpoint (shown in the API tab of this playground) with your WaveSpeedAI API key in the Authorization header. Submission returns a prediction ID. Poll the result endpoint starting around every 2 seconds, increase the interval for long-running tasks, and stop on any terminal status. The playground generates production-oriented Python, JavaScript, and cURL examples with timeouts, transient-error handling, and safe GET retries. Full request/response shape is documented at https://wavespeed.ai/docs/docs-api/nvidia/nvidia-nemotron-3-nano-omni-video.

How much does Nemotron 3 Nano Omni Video cost per run?

Nemotron 3 Nano Omni Video starts at $0.006 per run. That figure is the base price — the final charge scales with the parameters you set in the form (output size, length, count, references, or whatever knobs this model exposes), so a higher-quality or larger output costs more than a minimal one. The exact cost for your current input is shown live next to the Generate button before you submit, and the actual per-call charge is recorded on the prediction afterwards.

What inputs does Nemotron 3 Nano Omni Video accept?

Key inputs: `prompt`, `enable_sync_mode`, `max_tokens`, `reasoning_mode`, `system_prompt`, `temperature`. The full JSON schema (types, defaults, allowed values) is rendered above the Generate button and mirrored in the API reference at https://wavespeed.ai/docs/docs-api/nvidia/nvidia-nemotron-3-nano-omni-video.

How long does Nemotron 3 Nano Omni Video take to generate?

Median end-to-end generation time on WaveSpeedAI is around 8 seconds per request, based on recent successful runs. Queue time varies with global demand; live status is visible in the prediction record.

Can I use Nemotron 3 Nano Omni Video outputs commercially?

Commercial usage rights depend on the model's license, set by its provider (NVIDIA). The license summary appears on the model card above; see WaveSpeedAI's Terms of Service for platform-level conditions.

Nemotron 3 Nano Omni Video | AI Video Understanding API | WaveSpeedAI