agent-media 0.2.0

Python SDK for agent-media — generate AI UGC videos with talking heads, B-roll, and subtitles programmatically.

agent-media

Python SDK for agent-media — UGC video generation for developers.

Generate AI videos with realistic talking heads, B-roll, voiceover, animated subtitles, and music. One function call, finished video.

Install

pip install agent-media

Quick Start

from agent_media import AgentMedia

client = AgentMedia(api_key="ma_YOUR_KEY")

# Generate a video — blocks until complete, returns the URL
video = client.create_video(
    script="Have you tried building UGC content at scale? This API makes it trivial.",
    actor_slug="sofia",
    tone="energetic",
    style="hormozi",
    target_duration=10,
)
print(video["video_url"])
# → https://...mp4

Async

from agent_media.client import AsyncAgentMedia

async with AsyncAgentMedia(api_key="ma_YOUR_KEY") as client:
    video = await client.create_video(
        script="Your script here...",
        actor_slug="sofia",
        tone="confident",
    )
    print(video["video_url"])

API Reference

AgentMedia(api_key, base_url?, timeout?)

Parameter	Type	Required	Default
api_key	str	Yes	—
base_url	str	No	Production API
timeout	float	No	60.0

client.create_video(**params)

Generate a video and wait for completion. Returns { job_id, video_url, credits_deducted, duration }.

video = client.create_video(
    script="Your script (50-3000 chars)",
    actor_slug="sofia",           # use list_actors() to browse
    tone="energetic",             # energetic | calm | confident | dramatic
    music="chill",                # chill | energetic | corporate | dramatic | upbeat
    style="hormozi",              # 17 subtitle styles
    target_duration=10,           # 5 | 10 | 15 seconds
    aspect_ratio="9:16",          # 9:16 | 16:9 | 1:1
    allow_broll=True,             # include AI-generated B-roll
    template="product-review",    # optional template
    webhook_url="https://...",    # optional async callback
    timeout_seconds=600,          # max wait time (default 10 min)
)

client.submit_video(**params)

Submit without waiting. Returns { job_id, status, credits_deducted }.

client.submit_subtitle(**params)

Add subtitles to an existing video.

job = client.submit_subtitle(
    video_url="https://example.com/video.mp4",
    style="bold",
)

client.submit_product_review(**params)

Generate a product review video from a URL.

job = client.submit_product_review(
    product_url="https://example.com/product",
    angle="enthusiastic",
    actor_slug="sofia",
)

client.list_actors(limit?, offset?)

Browse available AI actors.

result = client.list_actors(limit=20)
for actor in result["actors"]:
    print(f"{actor['slug']} — {actor['name']} ({actor['gender']}, {actor['nationality']})")

client.get_video_status(job_id)

Check job progress.

status = client.get_video_status("job_abc123")
if status["status"] == "completed":
    print(status["video_url"])

Error Handling

from agent_media import AgentMedia, AgentMediaError

try:
    video = client.create_video(script="...")
except AgentMediaError as e:
    print(f"{e.code}: {e} (HTTP {e.status})")

Webhooks

Skip polling — pass webhook_url and agent-media will POST to it when the job completes or fails.

job = client.submit_video(
    script="...",
    actor_slug="sofia",
    webhook_url="https://example.com/webhooks/agent-media?secret=MY_TOKEN",
)

Payload on success:

{
  "job_id": "550e8400-e29b-41d4-a716-446655440000",
  "status": "completed",
  "video_url": "https://pub-xxx.r2.dev/videos/550e8400.mp4"
}

Payload on failure:

{
  "job_id": "550e8400-e29b-41d4-a716-446655440000",
  "status": "failed",
  "error_message": "Script exceeded maximum duration for selected actor."
}

Rules:

• Must be https:// (plain HTTP is rejected)
• Publicly reachable
• Max 2048 characters
• Retries on non-2xx: 3 attempts with exponential backoff (1 s, 4 s, 16 s)
• Query strings are preserved — append ?secret=MY_TOKEN to verify authenticity

Batch Generation

from concurrent.futures import ThreadPoolExecutor

scripts = ["Script 1...", "Script 2...", "Script 3..."]

with ThreadPoolExecutor(max_workers=3) as pool:
    videos = list(pool.map(
        lambda s: client.create_video(script=s, actor_slug="sofia"),
        scripts,
    ))

for v in videos:
    print(v["video_url"])

Related Packages

Package	Registry	Description
@agentmedia/sdk	npm	TypeScript SDK
@agentmedia/mcp-server	npm	MCP server for Claude Code, Cursor, Windsurf
agent-media-cli	npm	CLI tool
@agentmedia/schema	npm	Shared schema, types, Zod validation

Links

• Interactive API Docs
• OpenAPI Spec
• Website
• GitHub

License

Apache-2.0