Ingredient & Recipe Catalog

(Part 2 of 4 — same chapter in the PDF; split for the web site.)

Usage class other / core Summary Parse freeform user intents into structured video generation metadata using LLM-based analysis. Goes beyond simple prompt upsampling to provide rich metadata including style classification, multi-scene breakdown, camera work, audio planning, and detailed positive/negative prompts per scene. Llama-3.2-3B LLM Slug scrypted:llm:llama-3.2-3b Usage class other / core Summary The llama_3_2_3b ingredient provides access to the Llama-3.2-3B model, a compact language model with 3.2 billion parameters optimized for efficient text generation tasks including chat, completion, instruction following, and general-purpose text generation with fast response times. Qwen3-235B LLM Slug scrypted:llm:qwen3-235b Usage class other / core Summary The qwen3_235b ingredient provides access to the Qwen3-235B model, a large language model with 235 billion parameters optimized for various text generation tasks including chat, completion, instruction following, and complex reasoning. Code path (guess) ingredients/qwen3_235b/ Venice Uncensored LLM Slug scrypted:llm:venice-uncensored Usage class other / core Summary The venice_uncensored ingredient provides access to the Venice Uncensored model, which is specifically designed to handle content requests that most LLMs would reject, including adult content, controversial topics, and other sensitive material. Code path (guess) ingredients/venice_uncensored/ 101

Video Motion Type Lookup Slug scrypted:lookup:video-motion-type Usage class other / core Summary Provides motion type configuration lookup for video generation workflows, mapping motion type names to optimal target motion rates. Video Style Lookup Slug scrypted:lookup:video-style Usage class other / core Summary Provides style configuration lookup for video generation workflows, mapping style names to optimal target frame rates. FFmpeg Media Processor Slug scrypted:media:ffmpeg Usage class orchestration / media pipeline Summary Generic infrastructure ingredient for executing FFmpeg operations on video and image media. Handles downloading inputs, executing FFmpeg commands, managing temporary files, and uploading results to CDN. Code path (guess) ingredients/media_ffmpeg/ Media Stitcher Slug scrypted:media:stitch Usage class orchestration / media pipeline Summary Professional video stitching ingredient that combines multiple video and image segments with Adobe Premiere-style transitions. Handles timing, transitions, scaling, and composition. Code path (guess) ingredients/media_stitch/ 102

Store Media Slug scrypted:media:store Usage class orchestration / media pipeline Summary StoreMedia ingredient handler - downloads and stores media in our CDN. Store Base64 Images Slug scrypted:media:store-base64 Usage class orchestration / media pipeline Summary Store Base64 Images Ingredient. Takes base64-encoded images from Nova Canvas and uploads them to S3 ephemeral storage, returning S3 URIs that can be used by store_media. Video Orchestrator v1 Slug scrypted:orchestrator:video-v1 Usage class orchestration / media pipeline Summary Avoid re-running intent interpreter during execution JSON Extractor Slug scrypted:parser:json-extractor Usage class other / core Summary The JSON Extractor ingredient provides a robust, deterministic, synchronous service for extracting and repairing JSON from raw text output (typically from LLMs). It handles malformed JSON, markdown-wrapped JSON, multiple JSON objects, and various common LLM output issues. Code path (guess) ingredients/json_extractor/ Rule Assessment Parser Slug scrypted:parser:rule-assessment Usage class other / core 103

Summary The Rule Assessment Parser ingredient parses LLM responses containing rule-based assessments and returns structured assessment data. It uses the json_extractor ingredient vertically (embedded) to handle JSON extraction and repair, then validates and normalizes the assessment structure. Video Duration Planner Slug scrypted:planner:video-duration Usage class other / core Summary Intelligent scene duration planning that maps user-requested video durations (1-60s) to discrete Kling Video outputs (5s OR 10s) using optimal clip selection, trim, speed adjustment, and transition overlap calculations. Character Extraction Prompt Builder Slug scrypted:prompt:character-extraction Usage class other / core Summary Character Extraction Prompt Builder. Builds four Grok Imagine i2v prompts (T-pose front, side profile, back, face close-up) from vision analysis output, with optional character targeting and attributes. AWS Bedrock LLM Slug scrypted:provider:bedrock-llm Usage class AWS / Bedrock Summary The AWS Bedrock LLM ingredient provides a generic interface to AWS Bedrock language models, supporting multiple model families including Claude (Anthropic), Titan (Amazon), and Llama (Meta). This ingredient offers flexible authentication options and standardized response formats, making it easy to integrate LLM capabilities into recipes and workflows. Audio Generation - ElevenLabs Dialogue v3 Slug scrypted:recipe:audio-generation-elevenlabs-dialogue-v3 Usage class recipe-ingredient Summary Generate dialogue audio using ElevenLabs Dialogue v3 text-to-dialogue model via FAL.ai and store to CDN 104

HTTP path POST /generations/audio/elevenlabs-dialogue-v3 Character Extraction Slug scrypted:recipe:character-extraction Usage class recipe-ingredient Summary Extract character reference images (front T-pose, side, back, face) from a scene image using vision analysis and Grok Imagine i2v. Optional free-text to target a specific character (e.g. ’the man with black hair’). HTTP path POST /characters/extraction Content Moderation Pipeline Slug scrypted:recipe:content-moderation Usage class moderation / safety Summary End-to-end text and image moderation pipeline: content hashing, cache lookup, Bedrock Guardrails, routing, and optional permissive prompt rewriting Content Rewrite Recipe Slug scrypted:recipe:content-rewrite Usage class recipe-ingredient Summary Rewrites text content to address content policy violations while preserving original intent and structure Delula Prompts to Video - Google Veo 3.1 Slug scrypted:recipe:delula-prompts-to-video-google-veo-3-1 Usage class recipe-ingredient Summary Generate videos with optional audio from text prompts using Google Veo 3.1 Fast. Supports portrait, square, and landscape orientations at 1080p resolution. Audio generation is included in the video prompt when audio_prompt is provided. 105

HTTP path POST /generations/videos/delula-veo3-1-fast Delula Prompts to Video - NovaReel Slug scrypted:recipe:delula-prompts-to-video-novareel Usage class recipe-ingredient Summary Generate video from prompts using AWS Canvas for image generation and NovaReel for video generation, with optional orientation cropping and audio enhancement HTTP path POST /generations/videos/delula-novareel Time Series Forecast Slug scrypted:recipe:forecast Usage class recipe-ingredient Summary Generate probabilistic forecasts for time series data using the Sundial foundation model. Automatically handles normalization and expansion of input/output data. HTTP path POST /generations/forecasts Text to Image Generation Slug scrypted:recipe:generate-text-to-image Usage class recipe-ingredient Summary Generate images from text prompts with aspect ratio conversion and storage HTTP path POST /generate/image Text to Video Generation Slug scrypted:recipe:generate-text-to-video Usage class recipe-ingredient Summary Generate videos from text prompts with image-to-video conversion and storage 106

HTTP path POST /generate/video Video Orchestration Generation Slug scrypted:recipe:generate-video-orchestration Usage class recipe-ingredient Summary Generate multi-scene videos with intelligent orchestration, scene interpretation, and professional transitions HTTP path POST /generation/video/orchestration Video Orchestration V2 (Parallel Scenes) Slug scrypted:recipe:generate-video-orchestration-v2 Usage class orchestration / media pipeline Summary Parallel multi-scene video orchestration: intent analysis, per-scene image generation, video stitching with crossfade transitions, and MMAudio v2 soundtrack Image Generation - AWS Nova Canvas Slug scrypted:recipe:image-generation-aws-canvas Usage class recipe-ingredient Summary Generate images using AWS Nova Canvas with orientation support and store to CDN HTTP path POST /generations/images/aws-canvas Image Generation - FLUX 2 Pro Slug scrypted:recipe:image-generation-flux-2-pro Usage class recipe-ingredient Summary Generate images using Black Forest Labs FLUX 2 Pro text-to-image model via FAL.ai. Uses same orientation as Seedream-4 (portrait/landscape/square); landscape=1920x1080, portrait=1080x1920, square=1:1. Safety and format options, store to CDN. 107

HTTP path POST /generations/images/flux-2-pro Image Generation - Grok Imagine Image Slug scrypted:recipe:image-generation-grok-imagine-image Usage class recipe-ingredient Summary Generate images using xAI Grok Imagine Image text-to-image model via FAL.ai (xai/grok-imagine-image). Supports aspect_ratio, num_images, output_format (jpeg/png/webp). Store to CDN. HTTP path POST /generations/images/grok-imagine-image Image Generation - Nano-Banana Slug scrypted:recipe:image-generation-nano-banana Usage class recipe-ingredient Summary Generate images using nano-banana text-to-image model via FAL.ai with aspect ratio support and store to CDN HTTP path POST /generations/images/nano-banana Image Generation - Nano-Banana Edit Slug scrypted:recipe:image-generation-nano-banana-edit-image Usage class recipe-ingredient Summary Edit images using nano-banana image editing model via FAL.ai with multiple input images, aspect ratio support, and store to CDN HTTP path POST /generations/images/nano-banana-edit-image Image Generation - Nano-Banana Pro Edit Slug scrypted:recipe:image-generation-nano-banana-pro-edit-image Usage class recipe-ingredient 108

Summary Edit images using nano-banana-pro image editing model via FAL.ai with multiple input images, aspect ratio and resolution support, and store to CDN HTTP path POST /generations/images/nano-banana-pro-edit-image Image Generation - Nano-Banana Pro Slug scrypted:recipe:image-generation-nano-banana-pro-image Usage class recipe-ingredient Summary Generate images using nano-banana-pro text-to-image model via FAL.ai with aspect ratio and resolution support and store to CDN HTTP path POST /generations/images/nano-banana-pro-image Image Generation - Seedream 4 Slug scrypted:recipe:image-generation-seedream-4 Usage class recipe-ingredient Summary Generate images using ByteDance Seedream 4.0 text-to-image model via FAL.ai with aspect ratio support and store to CDN HTTP path POST /generations/images/seedream-4 Code path (guess) ingredients/seedream/ Moderation Text Assessment Slug scrypted:recipe:moderation-text-assessment Usage class recipe-ingredient Summary Two-stage content assessment: guardrail security check + rule-based LLM evaluation HTTP path POST /moderations/text/assessments 109

Structured Analysis (v2 - Parallel Processing) Slug scrypted:recipe:structured-analysis Usage class recipe-ingredient Summary Enhanced structured analysis with parallel processing of multiple analysis types Structured Image Analysis Slug scrypted:recipe:structured-image-analysis Usage class recipe-ingredient Summary Comprehensive structured analysis of images with style and motion classification HTTP path POST /analysis/image/structured Structured Video Analysis Slug scrypted:recipe:structured-video-analysis Usage class recipe-ingredient Summary Comprehensive structured analysis of videos with style and motion classification HTTP path POST /analysis/video/structured Text Surrounder Slug scrypted:recipe:surrounder Usage class converter / util Summary Wraps input text with configurable prefix and suffix strings for prompt formatting BGE-M3 Text Embeddings Slug scrypted:recipe:text-embeddings-bge-m3 Usage class recipe-ingredient 110

Summary Generate dense vector embeddings using BGE-M3 (BAAI General Embedding) model via Morpheus API Gateway for semantic search, RAG applications, and vector database indexing. HTTP path POST /generations/embeddings/bge-m3 Llama-3.2-3B Text Generation Slug scrypted:recipe:text-generation-llama-3-2-3b Usage class recipe-ingredient Summary Text generation using Llama-3.2-3B model (3.2B parameters) via Morpheus API Gateway for efficient, fast text generation HTTP path POST /generations/text/llama-3-2-3b Nova Pro Text Generation Slug scrypted:recipe:text-generation-nova-pro Usage class recipe-ingredient Summary Generic text generation using Amazon Nova Pro LLM with customizable system and user prompts HTTP path POST /generations/text/nova-pro Qwen3-235B Text Generation Slug scrypted:recipe:text-generation-qwen3-235b Usage class recipe-ingredient Summary Text generation using Qwen3-235B model (235B parameters) via Morpheus API Gateway for high-quality text generation HTTP path POST /generations/text/qwen3-235b Venice Uncensored Text Generation Slug scrypted:recipe:text-generation-venice-uncensored 111

Source: transcribed from the compiled Scrypted Network Design whitepaper PDF for web reading. Layout, figures, and pagination may differ from the PDF.