Imagen

# Gemini Imagen (Google AI Studio - free tier: 1500 req/day)
GEMINI_API_KEY=your_key

# fal.ai (pay-per-use, molti modelli)
FAL_KEY=your_key

# Replicate (pay-per-use, Flux + LoRA)
REPLICATE_API_TOKEN=your_token

Comandi

/imagen un gufo su un circuito cyberpunk
/imagen landscape at sunset --aspect 16:9
/imagen pixel art castle --backend flux
/imagen --output /path/to/file.png portrait photo
/imagen video: a dragon flying --backend fal
/imagen tts: "ciao mondo" --backend fal
/imagen music: jazz lounge background --backend fal

Wrapper Python

Gemini Imagen (immagini)

import sys
sys.path.insert(0, '.claude/skills/imagen')
from imagen import generate

path = generate("owl on circuit board, cyberpunk")
path = generate("pixel art castle", model="fast", aspect="16:9")
path = generate("logo design", output="/tmp/logo.png")

# Modelli: "fast", "standard", "ultra"
# Aspect: "1:1", "3:4", "4:3", "9:16", "16:9"

fal.ai (immagini, video, audio, TTS)

import sys
sys.path.insert(0, '.claude/skills/imagen')
from fal_client_wrapper import generate_image, generate_video, generate_audio, tts

# Immagine
path = generate_image("a cyberpunk owl", model="recraft", aspect="16:9")

# Video da testo
path = generate_video("a dragon flying over mountains", duration=5)

# Video da immagine
path = generate_video("the dragon breathes fire", image_path="/tmp/dragon.png")

# Musica
path = generate_music("epic orchestral battle, 30 seconds")

# Text-to-speech
path = tts("Buongiorno, oggi è una bella giornata.")

Replicate Flux (immagini + LoRA)

import sys
sys.path.insert(0, '.claude/skills/imagen')
from replicate_flux import generate, train_lora

# Base Flux
path = generate("a cat on a keyboard")

# Con LoRA custom
path = generate("my_character at a cafe", lora="my-character")

# Allena nuovo LoRA
model_id = train_lora(
    images_dir="/path/to/reference/images",
    name="my-character",
    trigger_word="my_character"
)

Intent Detection

args = "$ARGUMENTS".strip()
args_lower = args.lower()

# Detect media type
if any(w in args_lower for w in ["video:", "video di", "animazione"]):
    media_type = "video"
elif any(w in args_lower for w in ["tts:", "voce:", "leggi:", "say:"]):
    media_type = "tts"
elif any(w in args_lower for w in ["music:", "musica:", "audio:"]):
    media_type = "audio"