Runway Core Workflow A

Runway Core Workflow A

Overview

Advanced text-to-video generation: prompt engineering, model selection, parameter tuning, and batch generation.

Prerequisites

• Completed runway-hello-world

Instructions

Step 1: Model Selection

from runwayml import RunwayML
client = RunwayML()

# Available models:
# gen3a_turbo   — Fast, lower cost, good quality
# gen4_turbo    — Latest model, highest quality

task = client.image_to_video.create(
    model='gen4_turbo',
    prompt_text='A futuristic cityscape at night with flying cars and neon signs, cyberpunk aesthetic',
    duration=10,
    ratio='16:9',
)
result = task.wait_for_task_output()

Step 2: Prompt Engineering Tips

# Structure: Subject + Action + Setting + Style + Camera
prompts = [
    # Good: specific, visual, stylistic
    "A red fox walking through a snowy forest, soft winter light, documentary style, tracking shot",

    # Good: detailed motion and camera
    "Waves of golden wheat swaying in the wind, drone flyover, warm sunset, cinematic grain",

    # Bad: too abstract
    # "Something beautiful happening" — too vague
]

Step 3: Batch Generation

import asyncio

prompts = [
    "A butterfly emerging from a cocoon, macro lens, time-lapse, studio lighting",
    "Rain falling on a Tokyo street at night, reflections, neon, dolly zoom",
    "A chef preparing sushi in a traditional kitchen, close-up, warm lighting",
]

tasks = []
for prompt in prompts:
    task = client.image_to_video.create(
        model='gen3a_turbo',
        prompt_text=prompt,
        duration=5,
    )
    tasks.append(task)
    print(f"Queued: {task.id}")

# Wait for all
for task in tasks:
    result = task.wait_for_task_output()
    status = "OK" if result.status == "SUCCEEDED" else "FAILED"
    print(f"  {task.id}: {status}")

Step 4: Output Format Options

task = client.image_to_video.create(
    model='gen3a_turbo',
    prompt_text='Abstract paint mixing in slow motion, vibrant colors, black background',
    duration=5,
    ratio='9:16',      # Vertical for mobile/TikTok
    # ratio='16:9',    # Landscape for YouTube
    # ratio='1:1',     # Square for Instagram
)

Output

• Videos generated with optimal model selection
• Prompt engineering best practices applied
• Batch generation for multiple videos
• Output in various aspect ratios

Error Handling

Resources

• Runway API Documentation
• Input Parameters

Next Steps

Image-to-video: runway-core-workflow-b

Source: jeremylongshore/claude-code-plugins-plus-skills → skills/.curated/runway-core-workflow-a/SKILL.md

Also appears in: jeremylongshore/claude-code-plugins-plus-skills/plugins/saas-packs/runway-pack/skills/runway-core-workflow-a/SKILL.md