Your AI agent just wrote a script, generated an image, and composed a tweet — but what if it could also create a short video clip from a text description? Imagine an agent that generates product demos, social media snippets, explainer animations, or creative visualizations on the fly.
Agent Media Tools now offers an AI video generation API powered by Wan 2.5 via fal.ai. Submit a text prompt, wait a minute or two, and download the resulting MP4. It works from curl, Python, n8n, or directly from an AI agent via MCP — and since it's an async job, your agent can poll for status or register a webhook to get notified when the video is ready.
Text-to-image is everywhere now. Text-to-video is the next frontier — and it's finally practical for API-driven workflows. Here's why your agent should be able to generate video:
Unlike instant APIs that return results in milliseconds, AI video generation is compute-intensive. Each request takes 1–3 minutes depending on the duration and complexity. That's why the video generation API uses Agent Media Tools' async job queue:
job_id back immediatelyGET /api/jobs/:id to track progress (0–100%)Optionally, you can register a webhook and get a POST notification when generation finishes — no polling required.
| Property | Value |
|---|---|
| Endpoint | POST /api/agent/generate-video |
| Content-Type | application/json |
| Auth required | Yes — Bearer API key |
| Required field | prompt (text description of the video) |
| Optional fields | num_seconds (3–10, default 5), image_url (image-to-video) |
| Response | job_id — poll for result |
| Duration | 1–3 minutes typical |
| Output | MP4 video, downloadable for 1 hour |
| Pricing | 3 credits per second of generated video |
| Model | Wan 2.5 (via fal.ai) |
💡 Need an API key? Sign up free at agentmediatools.com and generate a key from your account settings. Free accounts can use credits to generate videos.
💡 Windows users: Run these from Git Bash or WSL — PowerShell's curl is an alias for Invoke-WebRequest. Or use Invoke-RestMethod with the same URL/body.
curl -X POST https://agentmediatools.com/api/agent/generate-video \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"prompt": "A serene mountain lake at sunset, with golden light reflecting on the water and pine trees swaying in the breeze",
"num_seconds": 5
}'
Response (immediate — job queued):
{
"success": true,
"job_id": "job_v001_a1b2c3d4e5",
"status": "queued",
"position": 1,
"estimated_seconds": 90
}
You'll get a job_id back. Now poll for the result:
curl -s https://agentmediatools.com/api/jobs/job_v001_a1b2c3d4e5 \ -H "Authorization: Bearer YOUR_API_KEY"
Polling response while processing:
{
"job_id": "job_v001_a1b2c3d4e5",
"status": "processing",
"progress": 65,
"progress_label": "Generating video frames..."
}
Polling response when complete:
{
"job_id": "job_v001_a1b2c3d4e5",
"status": "completed",
"result": {
"video_url": "https://agentmediatools.com/download/video_a1b2c3d4e5.mp4",
"duration_seconds": 5,
"size_bytes": 2843168,
"format": "mp4",
"prompt": "A serene mountain lake at sunset, with golden light reflecting on the water and pine trees swaying in the breeze"
},
"created_at": "2026-07-13T14:30:00Z",
"completed_at": "2026-07-13T14:31:25Z"
}
Download with curl:
curl -o sunset_lake.mp4 \ "https://agentmediatools.com/download/video_a1b2c3d4e5.mp4"
⏱️ Pro tip: Poll every 10–15 seconds with a short timeout. Add ?scroll=true for longer timeouts on slow connections. Or skip polling entirely by registering a webhook — we'll POST the result to your URL when generation finishes.
curl -X POST https://agentmediatools.com/api/agent/generate-video \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"prompt": "Cyberpunk city street at night, neon signs reflecting on wet pavement, flying cars in the distance, cinematic lighting",
"num_seconds": 10
}'
10-second videos cost 30 credits (3 credits × 10 seconds) and take 2–3 minutes to generate.
If you already have an image — maybe one generated by the image generation API — you can animate it by passing an image_url:
curl -X POST https://agentmediatools.com/api/agent/generate-video \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"prompt": "The character waves gently at the camera, smiling",
"image_url": "https://agentmediatools.com/download/artwork_xyz789.png",
"num_seconds": 4
}'
This creates a short animation that brings your still image to life — the model uses the image as the starting frame and generates motion based on the prompt.
import requests
import time
import sys
API_KEY = "YOUR_API_KEY"
BASE_URL = "https://agentmediatools.com"
# Step 1: Submit the job
headers = {
"Content-Type": "application/json",
"Authorization": f"Bearer {API_KEY}"
}
payload = {
"prompt": "A cat wearing a detective hat, walking through a noir-style rainy alley, dramatic shadows",
"num_seconds": 5
}
resp = requests.post(
f"{BASE_URL}/api/agent/generate-video",
headers=headers,
json=payload
)
data = resp.json()
job_id = data["job_id"]
print(f"🎬 Job submitted: {job_id}")
print(f" Estimated: {data.get('estimated_seconds', 90)} seconds")
# Step 2: Poll until complete
while True:
poll = requests.get(
f"{BASE_URL}/api/jobs/{job_id}",
headers=headers
)
status = poll.json()
if status["status"] == "completed":
print(f"\n✅ Video ready!")
video_url = status["result"]["video_url"]
print(f" Duration: {status['result']['duration_seconds']}s")
print(f" Size: {status['result']['size_bytes'] / 1024 / 1024:.1f} MB")
print(f" Download: {video_url}")
# Download the video
video_resp = requests.get(video_url)
filename = f"detective_cat_{status['result']['duration_seconds']}s.mp4"
with open(filename, "wb") as f:
f.write(video_resp.content)
print(f" Saved as: {filename}")
break
elif status["status"] == "failed":
print(f"\n❌ Failed: {status.get('error', 'Unknown error')}")
sys.exit(1)
else:
progress = status.get("progress", 0)
label = status.get("progress_label", "Queued...")
print(f" [{progress}%] {label}", end="\r")
time.sleep(10)
The upcoming agentmediatools Python package will make video generation simpler after registry publication. Until then, use the REST example above:
from agentmediatools import Client
client = Client(api_key="YOUR_API_KEY")
# Submit the job
job = client.generate_video(
prompt="Slow motion waves crashing on a tropical beach at golden hour",
num_seconds=5
)
# Wait for result (polls automatically)
result = job.wait(timeout=180)
# Download the video
result.download("beach_waves.mp4")
print(f"Downloaded {result.duration_seconds}s video to beach_waves.mp4")
If you use the Agent Media Tools MCP server, your agent can generate videos directly from natural language:
"Create a 5-second video of a futuristic city with flying cars and neon lights"
The MCP server exposes the generate_video tool. Claude will submit the job, poll for progress, and when the video is ready, it can show you the download link — all from a single instruction.
This is especially powerful in multi-step workflows:
"Generate an image of a sci-fi spaceship, then animate it with a slow flyby effect"
Claude will call generate_image first, grab the output URL, then call generate_video with that URL as the image_url parameter — all orchestrated automatically.
| Mode | Input | Best For |
|---|---|---|
| Text-to-video | Prompt only | Generating scenes from scratch — landscapes, abstract concepts, atmospheric clips |
| Image-to-video | Prompt + image URL | Animating existing artwork, adding motion to AI-generated images, content repurposing |
Both modes use the same endpoint and accept the same parameters — just include image_url when you want to animate from a starting frame.
Polling is simple, but webhooks are better. If you've registered a webhook URL in the Webhooks UI, video generation jobs automatically fire job.completed and job.failed events.
This means you can:
See the webhooks guide for setup instructions and receiver examples.
AI video generation is a premium, credits-based tool:
| Duration | Cost | Typical Wait |
|---|---|---|
| 3 seconds | 9 credits | ~45–60 seconds |
| 5 seconds | 15 credits | ~60–90 seconds |
| 10 seconds | 30 credits | ~2–3 minutes |
Purchase credit packs on the Pricing page:
Credits never expire and are shared across all your API keys. You can check your balance anytime with GET /api/agent/status or in your account settings.
An agent generates 10 short video clips from a list of product features, each 5 seconds long, then uploads them to a scheduling platform. Total time: ~15 minutes of generation for a week's worth of content.
A weekly newsletter agent writes the edition, generates a matching 3-second animated intro clip, attaches the video to the email, and sends — all without human involvement.
A creative director describes a scene in text, the agent generates a 5-second video, and the team reviews it before committing to full production. Iterate on prompts until the motion matches the vision.
An AI agent explains a complex topic (e.g., "how a combustion engine works") by generating a short animated diagram video from a text description, then presenting it alongside the written explanation.
| Feature | Agent Media Tools | Direct fal.ai / Wan API |
|---|---|---|
| Auth setup | Single API key | OAuth / complex auth flow |
| Job management | Built-in async queue | Must build your own |
| Progress polling | Built-in, returns % and label | Must implement yourself |
| Webhook on complete | Automatic (if configured) | Must build your own |
| MCP integration | Built-in tool | Not available |
| Billing | Credit packs, never expires | Pay-per-request + data egress |
| Unified ecosystem | Same key for 70+ tools | Video-only service |
The value isn't just the model — it's the pipeline infrastructure built around it. Async jobs, progress tracking, webhooks, and MCP integration turn a raw model API into something your agent can actually rely on.
⚡ Want to see it in action? Use the web UI at agentmediatools.com → Video Tools → ✨ AI Generate. Type a prompt, select a duration, and watch the progress bar fill up. No coding required to try it.
AI video generation is evolving fast. Here's what's on the roadmap:
All of these will be accessible through the same POST /api/agent/generate-video endpoint when they launch.
Sign up at agentmediatools.com, grab your API key, buy a small credit pack, and run the curl command above. In under two minutes you'll have a video generated from your description — no GPU needed, no model setup, no infrastructure.
Tags: video generation, text-to-video, AI video, Wan 2.5, AI agent API, async jobs, MCP video generation, generative AI, video API