Setvyn Local AI Workflow
Request beta access
Skip to content
Comic Story Director

Turn a story brief into structured, repeatable image-production scenes.

Organize projects, chapters, characters, looks, scenes, and panel prompts before sending reviewed production inputs to a configured ComfyUI workflow.

Project and chapter structure Optional OpenAI or Ollama assistance Configured ComfyUI handoff

Setvyn — Story DirectorLOCAL
Story brief Chapter rules Scene
The Lighthouse SignalChapter 01
Scene · Arrival at dusk
01
02
03
Panel 03 readyComfyUI queue
3 panelsScene composed
Projects Characters & looks Scenes & panels Model assistance ComfyUI handoff

From idea to panel input

Structure the chapter before repeating its context in every prompt.

A multi-panel sequence needs more than isolated prompt text. Setvyn keeps the chapter goal, continuity rules, selected characters, looks, locations, panel beats, prompt variants, and generation settings connected in one reviewable project.

Organize the sequence

Save projects and chapters, then arrange scenes and panels in the intended order.

Preserve the required context

Reuse character, look, location, person-count, and LoRA settings across selected panels.

Control every handoff

Review manual or assisted prompts before sending one panel or a selected group to ComfyUI.

01 · Structure

Give every chapter a reusable production context.

Create or load a saved comic project, divide it into chapters, and keep each chapter's story brief, continuity rules, characters, panel sequence, and generation settings together.

  • Saved projects and chapters
  • Chapter names and panel counts
  • Story brief and chapter continuity rules
  • Create, load, update, and continue a chapter
  • Local autosave feedback while editing
Project navigatorAutosaved
The Lighthouse SignalProject · 2 chapters
Chapter 01 · Arrival6 panels
Chapter 02 · The Fog4 panels
Load chapter Continue

Local project structure with autosave feedback. This is not collaborative cloud editing, source-control history, real-time multi-user editing, or unlimited version restoration.

02 · Cast

Keep identity, appearance, and chapter-specific looks explicit.

Define the characters used in a chapter, attach their identity prompts, notes, person types, and one or more looks, then choose the relevant character and look for each panel.

  • Multiple characters per chapter
  • Identity prompts and optional notes
  • Reusable outfits or visual looks per character
  • Custom person types and automatic count tokens
  • Character- and look-level LoRA choices with available trigger metadata
Character context
A AriLead · person type: adult 2 looks
Identity prompt short dark hair, weathered scarf, calm expression
Field JacketLook · LoRA: field-v2
Night LookLook · trigger available

These settings help prepare consistent input but cannot guarantee identical characters across generated images. Final consistency depends on the models, LoRAs, prompts, workflows, seeds, and user review.

03 · Direct

Break a chapter into scenes, beats, and panel-level decisions.

Create scenes, add and reorder panels, select the characters present, assign their looks, describe the panel beat, and keep manual and assisted prompt output available for comparison.

  • Scene and panel creation, removal, and reordering
  • Panel brief and location context
  • Selected character and look per panel
  • Manual prompt and assisted prompt modes
  • Panel seeds, selection state, and per-panel generation overrides
  • Send the current panel or a selected panel group to the next step
01 Wide shot · lighthouse at duskBeat · establishing seed 1041
02 Signal activates · fog increasesBeat · rising tension seed 1042
03 Close detail · control panel respondsBeat · payoff Selected

Panels are text-only production beats. Setvyn does not perform automatic page layout, speech balloons, lettering, print preparation, or finished sequential artwork.

04 · Refine

Ask for structured panel prompts while preserving chapter rules.

When configured, an OpenAI or Ollama backend can refine the current panel, selected panels, a scene, or a batch using the story brief, chapter rules, character context, and target panel count.

  • Configured OpenAI or Ollama backend profiles
  • Current-panel, selected-panel, scene, or batch requests
  • Optional generation of panel input text
  • Short, medium, or detailed panel and location output
  • Option to enforce the exact selected character count
  • Adjustable model parameters such as maximum tokens and sampling controls
Panel assist
Panel brief Ari reaches the control panel as the fog thickens; tight framing on hands.
Structured prompt draftReview before use close-up, single figure, weathered hands on lit control panel, dense fog, cool rim light, muted teal palette, 3:2
Detail: mediumCount enforced: 1Backend: Ollama

LLM output is a draft for review and can be inaccurate or inconsistent. Setvyn does not guarantee continuity, correct anatomy, perfect prompts, or automatic compliance with your intent.

Production overrides

Keep shared workflow settings and panel-specific choices understandable.

The Story Director can read LoRA information from the configured ComfyUI workflow, combine it with chapter choices, and add character, look, or panel overrides for the panel being prepared.

Each layer remains visible so the creator can inspect what will be sent. A panel can also be configured to ignore workflow LoRAs where that option is supported.

1 Workflow base from ComfyUI
2 Chapter overrides continuity rules
3 Character & look choices per cast
4 Panel overrides final layer

Not every ComfyUI workflow or custom node can be parsed. Availability depends on the configured workflow and supported LoRA manager structure.

05 · Queue

Send prepared panel input to the ComfyUI workflow you configure.

Choose a panel, review its final prompt and negative prompt, select a resolution profile, apply the effective LoRA settings, and submit it to the configured ComfyUI server and workflow.

  • Configured ComfyUI server connection
  • API and UI workflow file configuration
  • Reusable or custom resolution profiles
  • Negative prompt and panel prompt preview
  • Workflow model and LoRA context where supported
  • Returned prompt ID, queue number, and recent chapter queue entries
ComfyUI handoffLocal server configured
PanelPanel 03
ResolutionPortrait 1216 × 1824
Panel prompt close-up, single figure, control panel, dense fog, cool rim light
Negative prompt extra fingers, text, watermark, low detail
Submitted to workflow Queue position 2prompt accepted

Setvyn does not render the image itself. ComfyUI performs generation through your configured workflow and environment. Values shown are placeholders—no real server URL, endpoint, workflow path, model path, API response, prompt ID, or private node data.

Know where the data goes

Keep the chapter local and connect external services only for selected actions.

Projects, chapters, scenes, panel drafts, character context, and workflow configuration stay in the local Setvyn project. Sending a panel to ComfyUI transfers the required generation input to the configured ComfyUI server. Using an external OpenAI backend sends the required text to that provider; a configured Ollama backend can remain local.

Local project

Story structure, characters, looks, scenes, panel settings, saved prompts, and queue references.

Configured LLM

Only used for the selected refinement action; external-provider behavior depends on the backend chosen.

Configured ComfyUI

Receives the reviewed prompt and supported workflow inputs needed to queue the selected panel.

Not every operation is offline. Setvyn does not expose tokens, endpoints, local paths, model names, workflow JSON, or customer prompts.

Use the same production context before and after the panel draft.

Character Library

Prepare reusable character identity and look context for wider Setvyn workflows.

Prompt Tools

Filter, split, or direct prompt text before another handoff.

Learn more
Selection

Review generated image results and choose the assets that continue into production.

Operations & Training

Inspect supported local training tasks and model status used by quality-control workflows.

Choose the story workflow depth that fits your production.

Story Director, model assistance, ComfyUI, LoRA, and batch capabilities vary by Setvyn edition and local configuration.

Compare Setvyn editions

FAQ

Comic Story Director FAQ

Does Setvyn generate a finished comic automatically?

No. It structures projects, chapters, scenes, characters, panels, and prompts, then supports a reviewed handoff to the configured image workflow.

Can I reuse the same character across panels?

Yes. Character identity, person type, looks, prompts, and supported LoRA choices can be reused across selected panels. The final visual consistency still depends on the configured generation workflow and user review.

Do I need an external LLM?

No external provider is required for manual panel planning. Optional model assistance can use a configured OpenAI or Ollama backend.

Does the Story Director run ComfyUI itself?

No. It prepares and submits supported panel inputs to the ComfyUI server and workflow configured by the user.

Can I send several panels together?

The workflow supports selecting panels for grouped LLM preparation and ComfyUI submission where the configured operation allows it. Every panel should still be reviewed.

Does it create speech balloons or page layouts?

No page-layout, lettering, or speech-balloon workflow is promised on this page.

Direct the sequence before generation

Keep chapter rules, characters, panels, and handoff settings connected.

Move from a story brief to reviewed panel inputs without rebuilding the production context at every scene.

Request beta access Explore all features