Shotstack Workflows

Provides a drag-and-drop canvas interface for constructing generative AI media pipelines without code. Users connect pre-built nodes representing media operations (generation, editing, composition) with visual connectors that define data flow and execution order. The builder compiles workflows into executable DAGs (directed acyclic graphs) that handle dependency resolution and parallel execution where possible.

Embeds pre-configured nodes that interface with generative AI models for text-to-image, text-to-video, image editing, and style transfer operations. Each node abstracts API calls to underlying AI providers (likely including Shotstack's own rendering engine and third-party models) with parameter mapping, prompt engineering templates, and result caching. Nodes handle model selection, parameter validation, and error recovery automatically.

Maintains version history of workflows, allowing users to save snapshots at key points and revert to previous versions if needed. Each version captures the complete workflow definition (nodes, connections, parameters) with metadata (timestamp, author, change description). Supports comparing versions to identify changes and rolling back to any previous version without losing current work. Versions can be tagged for easy reference (e.g., 'production-v1', 'testing').

Allows users to save workflows as reusable templates with parameterized inputs (e.g., {{videoTitle}}, {{brandColor}}, {{duration}}). Templates support variable substitution at runtime, enabling batch processing and personalization without rebuilding workflows. Parameters are validated against type schemas and can be provided via API calls, CSV uploads, or manual input, with support for conditional parameter visibility based on workflow state.

Exposes REST API endpoints that trigger workflow execution with JSON payloads and deliver results via configurable webhooks. Workflows can be invoked synchronously (waiting for completion) or asynchronously (returning a job ID for polling). Results are posted to user-specified webhook URLs with signed payloads for security, supporting retry logic with exponential backoff for failed deliveries. Integrates with external systems (Zapier, Make, custom applications) via standard HTTP callbacks.

Includes nodes for compositing multiple media assets (images, videos, text, effects) onto a timeline with frame-accurate positioning, timing, and layering. Supports keyframe animation for properties like position, scale, opacity, and rotation. Timeline-based editing allows users to define when each element appears, how long it displays, and how it transitions. Composition nodes handle rendering optimization by pre-calculating frame sequences and managing memory efficiently.

Enables workflows to branch based on runtime conditions (e.g., if image generation succeeds, proceed to composition; otherwise, use fallback image). Conditions evaluate against workflow state, node outputs, or external data using simple rule engines (e.g., if {{quality}} > 0.8, then use high-res output). Supports multiple branches with fallback paths and error handling, allowing workflows to adapt to different inputs or execution outcomes without requiring separate workflow definitions.

Tracks workflow execution in real-time with detailed logs capturing each node's input, output, duration, and status. Provides a dashboard showing execution history, performance metrics, and error details. Logs are stored for audit trails and debugging, with filtering and search capabilities to identify issues. Execution metrics include node-level timing, resource usage, and success/failure rates, enabling optimization of slow workflows.

Accepts batch input files (CSV or JSON) containing multiple sets of workflow parameters and executes the workflow for each row/object. Processes batches asynchronously with configurable concurrency limits to manage resource usage. Results are aggregated and exported in bulk formats (CSV, JSON, or ZIP archives containing media files). Supports progress tracking and partial failure recovery, allowing batches to continue processing even if individual items fail.

Allows developers to create custom nodes that extend workflow capabilities beyond built-in operations. Custom nodes are defined via configuration (JSON schema for inputs/outputs) or code (JavaScript/Python functions) and integrate into the workflow canvas like native nodes. The framework handles node lifecycle (initialization, execution, cleanup), error handling, and result propagation. Custom nodes can wrap external APIs, implement business logic, or perform specialized media operations.

Provides a centralized repository for storing and organizing media assets (images, videos, audio, fonts, templates) used in workflows. Assets can be tagged, versioned, and searched, with support for organizing into folders or collections. Workflows reference assets by ID or path, enabling reuse across multiple workflows without duplicating files. Integration with cloud storage (S3, Google Cloud Storage) allows importing external assets and exporting results directly to cloud buckets.