7 Modalities.
One Pipeline.

from pydantic import BaseModel
from google import genai
from google.genai import types

class VideoEvent(BaseModel):
    start_time: str  # MM:SS
    end_time: str    # MM:SS
    event: str
    visual_details: str

client = genai.Client()
video_bytes = open("demo/assets/00_source/omni_source.mp4", "rb").read()

response = client.models.generate_content(
    model="gemini-3.8-flash",
    contents=[
        types.Part.from_bytes(data=video_bytes, mime_type="video/mp4"),
        "Extract every visual edit, brand logo, and natural ad-break timestamp (MM:SS).",
    ],
    config=types.GenerateContentConfig(
        response_mime_type="application/json",
        response_schema=list[VideoEvent],
    ),
)

from pydantic import BaseModel
from google import genai
from google.genai import types

class CaptionSegment(BaseModel):
    speaker: str
    start_time: str  # MM:SS
    end_time: str    # MM:SS
    text: str
    tone: str

client = genai.Client()
audio_bytes = open("demo/assets/00_source/omni_audio.mp3", "rb").read()

verbatim = client.models.generate_content(
    model="gemini-3.5-transcribe",
    contents=[types.Part.from_bytes(data=audio_bytes, mime_type="audio/mp3"), "Transcribe verbatim."],
)

sdh = client.models.generate_content(
    model="gemini-3.8-flash",
    contents=[types.Part.from_bytes(data=audio_bytes, mime_type="audio/mp3"), "Transcribe with MM:SS and tone."],
    config=types.GenerateContentConfig(
        response_mime_type="application/json",
        response_schema=list[CaptionSegment],
    ),
)

from pydantic import BaseModel
from google import genai
from google.genai import types

class AdDecision(BaseModel):
    matched_campaign_id: str
    brand: str
    scte35_insertion_point_mm_ss: str
    dynamic_ctv_video_prompt: str
    dynamic_voiceover_script: str

client = genai.Client()

emb = client.models.embed_content(
    model="gemini-embedding-2",
    contents="Host holds a BROX COFFEE cup in a wood-slat podcast studio.",
)

decision = client.models.generate_content(
    model="gemini-4-argon",
    contents="Match this clip to the Brox Coffee brief and output a CTV ad plan.",
    config=types.GenerateContentConfig(
        response_mime_type="application/json",
        response_schema=AdDecision,
    ),
)

from google import genai
from google.genai import types

client = genai.Client()
keyframe_bytes = open("demo/assets/04_image_gen/raw_keyframe.png", "rb").read()

response = client.models.generate_content(
    model="gemini-nano-banana-2.1",
    contents=[
        types.Part.from_bytes(data=keyframe_bytes, mime_type="image/png"),
        "Transform this frame into a high-contrast 16:9 CTV thumbnail: keep the host and "
        "BROX COFFEE cup identical, add cyan rim lighting, a stormy ocean vortex outside "
        "a studio window behind him, and bold white text 'THE OMNI SHIFT' in the upper right.",
    ],
    config=types.GenerateContentConfig(
        response_modalities=["TEXT", "IMAGE"],
    ),
)

from pydantic import BaseModel
from google import genai
from google.genai import types

class BoundingBox(BaseModel):
    label: str
    box_2d: list[int]  # [ymin, xmin, ymax, xmax] normalized 0..1000

client = genai.Client()
keyframe_bytes = open("demo/assets/04_image_gen/raw_keyframe.png", "rb").read()

response = client.models.generate_content(
    model="gemini-3.8-flash",
    contents=[
        types.Part.from_bytes(data=keyframe_bytes, mime_type="image/png"),
        "Detect host face, BROX COFFEE cup, podcast microphone, and tropical plant. "
        "Return 2D bounding boxes as [ymin, xmin, ymax, xmax] scaled 0 to 1000.",
    ],
    config=types.GenerateContentConfig(
        response_mime_type="application/json",
        response_schema=list[BoundingBox],
    ),
)

import base64
from google import genai

client = genai.Client(vertexai=True, project="ckbrox13", location="global")
img_b64 = base64.b64encode(open("demo/assets/00_source/advertiser_espresso_can.png", "rb").read()).decode()

# 1. Create 16:9 CTV Video Ad from Static Product Photo
v1 = client.interactions.create(
    model="gemini-omni-1.1-flash-preview",
    input=[
        {"type": "image", "data": img_b64, "mime_type": "image/png"},
        {"type": "text", "text": "BROX COFFEE cup and cold-brew can rotating slowly with rising steam."},
    ],
    response_format={"type": "video", "aspect_ratio": "16:9", "resolution": "720p"},
)

# 2. Conversationally Edit Video Background into the Host's Wood-Slat Studio
v2 = client.interactions.create(
    model="gemini-omni-1.1-flash-preview",
    previous_interaction_id=v1.id,
    input="Place the BROX COFFEE cup and can in a modern wood-slat podcast studio next to a Shure SM7B mic.",
    response_format={"type": "video", "aspect_ratio": "16:9", "resolution": "720p"},
)

from google import genai
from google.genai import types

client = genai.Client(vertexai=True, project="ckbrox13", location="global")

# 1. Expressive TTS Voiceover with Acting Cues (gemini-3.8-flash-tts)
tts_res = client.models.generate_content(
    model="gemini-3.8-flash-tts",
    contents=(
        "From Blue Bottle to Starbucks to Brox Coffee—when you are building the future of AI, "
        "only one roast keeps up. <sigh> Brox Coffee. Freshly roasted in San Francisco and New York."
    ),
    config=types.GenerateContentConfig(
        response_modalities=["AUDIO"],
        speech_config={"voice_config": {"prebuilt_voice_config": {"voice_name": "Aoede"}}},
    ),
)

# 2. Backing Music Generation (lyria-3-clip-preview)
music_res = genai.Client().interactions.create(
    model="lyria-3-clip-preview",
    input="Upbeat minimal electronic creator podcast beat, 114 BPM. Instrumental only, no vocals.",
)