Skip to content

Text and authoring models

Text models write or revise text. They do not render the image or video described in that text. Choose the authoring node’s model separately from the image or video model that will consume its result.

The tables below enumerate the current selector, including a saved-node-only legacy value. Copilot and Video Analysis have their own configured models; changing an ordinary text node does not change those services.

The current default for the Claude text-model selector. Use the generated default below for new authoring nodes; a saved node can retain another model.

Setting Value
Name Claude Opus 5.5
Availability available
Default for new nodes yes

An offered Claude authoring option. This guide has no controlled Fable-versus-Opus script-quality comparison; judge the generated scene plan against the brief.

Setting Value
Name Claude Fable 5.1
Availability available
Default for new nodes no

An offered Claude option for existing or new authoring setups. It remains a distinct model record from Opus 5.5.

Setting Value
Name Claude Opus 5
Availability available
Default for new nodes no

An offered Claude option. The Studio copilot uses a fixed model separately from the selector on an authoring node; changing a node does not change the copilot.

Setting Value
Name Claude Sonnet 5
Availability available
Default for new nodes no

An older offered option. An older model remaining selectable is different from a retired model.

Setting Value
Name Claude Opus 4.8
Availability available
Default for new nodes no

Deprecated for new selection. A saved node can retain this legacy value; do not use this entry as a recommendation for a new project.

Setting Value
Name Claude Sonnet 4.6
Availability legacy: existing nodes only
Default for new nodes no

The OpenAI option exposed in the text selector. A different model choice may change wording and structure, so review the complete output before downstream generation.

Setting Value
Name gpt-5.4-mini
Availability available
Default for new nodes no

Give the writer the goal, the intended audience, the facts it must preserve and the shape of the required output. Separate dialogue from visual action. Name fixed constraints explicitly rather than asking for “something cinematic”.

For Prompt for the model, select the target video family and supply a scene such as: “One gardener opens a greenhouse door, notices a wilted plant, and waters it. Start before the door opens. Keep the gardener recognizable. No character dialogue.” Read the resulting text, check the opening state, and then choose the separate Video model. A polished prompt does not prove the clip will satisfy it.

A writer can omit part of the brief, introduce a new character, put a reveal in the first frame, or give narration to an on-screen actor. Check the entire plan, including sound and references. Preserve omitted user requirements through a revision rather than silently shortening the brief.

The current model list and legacy policy are verified from the shared selector contract. This page does not claim a comparative writing benchmark across these seven model records. The dated video observations on model pages concern specific downstream renders, not proof of a text model’s general quality. See action and timing and narration.

Reviewed 14 October 2026. Author documentation: Claude model overview and GPT-5.4 mini. The Studio selector is authoritative for what is offered here; the author’s full catalog is larger. Recheck on model or authoring-contract changes.