Skip to main content
Send a prompt through REST or MCP to create a video. New video jobs default to GVA with standard effort. Read engine discovery for your account’s available engines, supported efforts and default.

Engine and effort

Motion supports Light, Medium and High effort. Medium is the default. Effort affects runtime and credit use; the task and generated media also affect the final cost. To choose Motion and High effort explicitly:
In an MCP client, ask your agent: “Use Motion with high effort.” It can send the same fields to create_video.

Discover available engines

MCP clients call list_engines. Each result reports engine, name, efforts, default_effort, default, and whether templates are supported. Read this catalog rather than hardcoding every engine; future engines can use the same request surface. Engine discovery includes only engines your account can use. GVA is available to signed-in accounts. Motion Legacy is available only to eligible existing accounts and administrators. API keys, OAuth scopes, and credits do not change that eligibility. An explicit unavailable engine returns HTTP 403 with engine_access_required. GVA is the default for new jobs without an explicit engine, including requests using the deprecated REST mode field. Your existing API key or MCP connection continues to work. Clients that cache engine choices should refresh discovery. Existing projects keep their engine. Access is checked on creation and every follow-up, including credit-pause resumes. Owners can still read existing outputs.

GVA

Standard and Extended are effort levels of the same gva engine. Use Standard for everyday work or Extended for more demanding tasks: For Standard:
For Extended:
In ChatGPT or another MCP client, ask “Use GVA Standard” or “Use GVA Extended.” The agent sends the matching engine and effort fields. You can also omit engine and send effort: "extended" when creating a new GVA job. Motion Legacy eligibility is separate from GVA access. Read engine discovery before choosing Motion explicitly.

Engine-specific templates

List templates for your selected engine:
Or call MCP list_templates({"engine":"motion"}). Without an engine, discovery uses your account’s default. Only published templates available to your account appear; an empty list means none are currently available. Use a returned identifier as template when creating a job. Templates seed the editable project before the first turn and require Pro or Max. Choose a template from the same engine’s catalog. Personal Brand templates require GVA Standard; select effort: "standard" when using one.

Revise and resume

Follow-ups keep the project’s engine, effort, and template. Omit effort to keep the current choice, or set it for the next turn. A GVA job can switch between standard and extended; an existing Motion job stays on Motion and uses light, medium, or high:
Send the request to POST /sessions/{job_id}/followups, or use MCP create_followup with the same job_id. The job must be completed or waiting on user input. After topping up a credit-paused job, continue that same job. Create a new job to change engines or templates. Responses include the effective engine, effort, and template; new engine-backed jobs require a downloadable export before reporting completed. Older session jobs can report completed while an export is pending. For every job, check output.status === "completed" and a non-null output.download_url before downloading.

Existing integrations

Endpoint URLs, API-key prefixes, OAuth scopes, tool names, and job IDs remain valid. Existing projects retain their engine. New requests that omit engine use GVA. Existing Motion chats, including pre-launch chats without engine metadata, continue using Motion on follow-up. The following deprecated inputs are supported by REST. MCP clients must use the published engine and effort fields: engine: "motion" selects Motion. MCP schemas do not expose mode. Earlier GVA effort values remain accepted by MCP as effort aliases. New integrations should use engine and effort. Read the effective selection from the response rather than assuming all saved projects use GVA.