Scenario Blender Previs Storyboard logo

Scenario Blender Previs Storyboard

Organization
scenario-labs
scenario-blender-previs-storyboard

Use when turning a script or story into previs, layout, storyboards or an animatic in Blender: beats to shot list, camera and lens per shot, staging, camera height, 180-degree rule, screen direction, shot sizes, when to cut, camera moves, timing stills in the Video Sequencer, markers bound to cameras, Story Tools, Grease Pencil or 2D/3D hybrid boards. Also when an animatic shows the wrong frame, blocking keys are not stepped, Rigify operators are missing, or viewport render fails headless.

Overview

Publisherscenario-labs
Repositoryskills
Skill namescenario-blender-previs-storyboard
Stars
970
Forks
114
Bundled files
5
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 5 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by scenario-labs on GitHub. Read the source before you install it.

Installation

Install the Scenario Blender Previs Storyboard AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/scenario-labs/skills/tree/main/skills/dcc/blender/scenario-blender-previs-storyboard
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/scenario-labs/skills.git /tmp/skills
mkdir -p .claude/skills
cp -r /tmp/skills/skills/dcc/blender/scenario-blender-previs-storyboard .claude/skills/scenario-blender-previs-storyboard
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Scenario Blender Previs Storyboard in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Scenario Blender Previs Storyboard on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Scenario Blender Previs Storyboard is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Previs, layout and storyboarding

Expert previs conveys an idea well enough to be judged, as cheaply as possible: one fact per beat, every shot designed for what the audience must feel, timing found in the edit from stills. The camera is the audience's body; where it stands, how high and through which lens decides the emotion, so every non-neutral choice needs a stated reason. If a sibling skill named here is missing from your available skills, ask the user to install it (npx skills add scenario-labs/skills --skill <name>); unattended, proceed from tool schemas and flag the gap.

REQUIRED BACKGROUND: scenario-blender-expert (execution channel, review loop, 5.2 API changes).

Toolkit: scripts/bx_previs.py (tested on 5.2.1): import sys; sys.path += ["<this skill>/scripts", "<scenario-blender-expert>/scripts"]; import bx_previs as P.

Which skill. This one: story previs and animatics judged by a director and an editor. text-image-to-blender-blockout instead when the playblast guides a render-to-real video model (Seedance): saturated color identities, no lens jump on a cut, eased moves, a hand-off sheet; an approved animatic that must become a video guide is rebuilt there. Shared: metric Z-up, roots at the feet, one camera per shot bound to a marker, to_track_quat('-Z','Y') look-at. Strokes: scenario-blender-grease-pencil. Animation after layout: scenario-blender-animation. Real rigs: scenario-blender-rigging.

Stance (the expert delta)

  • Hjalti Hjalmarsson: write the beats-of-information list before any shot. One fact per line, because a viewer takes in one thing at a time. Group consecutive beats into shots; a new treatment re-orders the same facts. A missing beat means the idea never gets a fair test.
  • Hjalti: theater first, then cinematic; shots cost time. A locked-off version judges the acting; the same beats told in shots ran about 2x (up to 3x) because every cut must be read (13/22/25 s vs 47/40 s). The oversimplified version is always too short and hides logistics (a character cannot cross the room in 10 frames).
  • Charge recipe (Hjalti): treat every frame as a storyboard drawing. 10 frames per shot, keys CONSTANT, an A/B alternative on the next frame, the winner on the decade frame, stills exported with versioned names, timing done in a separate edit.
  • Hjalti: when in doubt, eye height; lock the lens early by A/B. Two characters: split their eye heights. Low or high is a statement. Poses built for one lens do not survive a lens change.
  • Dillon Gu: composition and focal length are separate decisions, and cuts can kill a mood. The audience stands where the camera is. Leaving the space removes the danger; two inanimate inserts in a row lower tension.
  • Cheat for the camera, but key before you cheat (Hjalti, Charge, Renato Roldan). Resize, float, swap props per shot; in a shared timeline an unkeyed cheat breaks the earlier shots.
  • Previs is the cheapest version that still conveys the idea (Hjalti, Charge). Animate only shots whose camera move is the storytelling; everything else is one still, push-ins faked on the still in the edit. Grays in Workbench, no mood lighting, mannequins before rigs. P.audit refuses a camera moving inside a still shot, keys between block starts and a non-Workbench engine.
  • After each cut the eye must find one thing (Hjalti). Restage a character in front of a competing subject; keep the angle that preserves the emotion (he refused a side cut that turned Rex into a silhouette).

Establish first

InputChangesDefault when silent
Formatevery composition (Charge: lock it first)1920x1080, 24 fps, stills at 50 %
Treatmentshot count, runtimetheater pass, then cinematic
Methodagent has no tablet3D previs with mannequins; drawn boards only on provided art; Rigify proxies only on request (P.rigify_proxy, Rigify is off under --factory-startup: P.ensure_rigify() first)
Mood or director stylelens, height, motion per shotneutral grammar, deviations stated per shot
Runtime budgetdurationsabout 1.6 to 1.9 s per beat for a cinematic cut (derived from Hjalti's slide: 25 beats, 8 shots, 40 to 47 s)
Journeysscreen directionout left to right, back right to left (Charge)
Deliverableoutputsshot list, contact sheet, floor plan, animatic MP4 clean plus burn-in, versioned .blend

Workflow

  1. Script to beats. Asset list, then numbered beats, one fact each; mark every beat you add. GATE: P.shot_list(BEATS, SHOTS) raises on a missing, duplicated or reordered beat.
  2. Set and proxies. P.setup_scene (format, Workbench, grays, Standard view transform); P.prop for real-scale set pieces; P.mannequin for characters (root at the feet with eye_height, four pivot controls as the layout subset). Grays, one pop color for the hero (Renato: red). GATE: scale reads next to a door (2.1 m) and counter (1.0 m).
  3. Theater pass. One locked-off camera (cam shared, fit="all", 35 mm, mean eye height) framing the action's extremes; one block per beat group. GATE: in the P.audit rows every subject's head_in and feet_in is true at every block; theater edit timed.
  4. Cinematic plan. List several openings, re-order beats, group them into shots with intent, size, lens, height, dur. P.floor_plan with the action line and journey paths before judging cameras. GATE: one side of the line; journeys follow the direction rule.
  5. Blocking. For each block: sc.frame_set(f) (holds the previous pose), set only what changes, P.key_block(all_movables, f); cheats go through the same keys; then P.make_stepped. GATE: P.stepped_report == {}; the cheat check at neighboring shots.
  6. Cameras. P.build_shots places each camera from size, lens and eye height (or explicit loc/aim), sets an opaque passepartout (Charge: see only the composition; P.passepartout(sc) for other cameras) and binds a marker at its first frame. P.ab_lens on key shots before posing further. GATE: P.audit(...)["problems"] == [].
  7. Look. P.contact_sheet and open it. Characters merging with a dim set: P.fog_for_shot(sc, t, chars) (Hjalti: depth fog for readability, not lighting; behind the subjects, keyed to that shot), P.values_at hero vs backdrop. GATE: the visual list under Quality gates (in S6 only the pictures caught the occlusions and the cut-off wide).
  8. Edit. P.render_stills, then P.animatic (image strips held dur seconds, push on stills, live shots as scene strips, sounds, burn_in). One continuous action from several angles (Pablo, Story Tools): a live shot plus cover=True shots become one scene strip soft-split per camera; retime with P.retime_cut (the marker follows), never by moving keys. GATE: runtime within budget; frames pulled from the MP4 match the cut list.
  9. Iterate in a new scene pair (02_blend/02_edit) and new version numbers; nothing overwritten.

Numbers

ValueRelative toSource
10 frames per shot block, 20 only when justified; starts on 10, 20, 30previs timelineCharge
Shots named in tens (sh010, sh020)markersHjalti on screen
35 mm theater frame; about 60 mm for an isolating key shot36 mm sensorHjalti
F-stop 5.6, clip 0.01 to 200 m on the camera rigHjalti's rigHjalti
13 / 22 / 25 s theater vs 47 / 40 s cinematicsame scriptHjalti slide 00:46:01
3 poses for an anger burst (anticipation, overshoot, settle)one beatHjalti
6 fps boarding, retimed in the editboard sceneBouncyBrain
Eye tolerance 0.15 m; drawing breaks past 70 degrees from cameraaudit thresholds[added]
Visible height = subject height x 4 / 2 / 1.15 / 0.8 / 0.6 / 0.4 / 0.25 / 0.12ews, wide, full, mws, medium, mcu, cu, ecu[added]
Distance = visible height x lens / sensor height (20.25 mm at 16:9)lens swap at equal size: d2 = d1 x f2 / f1[added]
Handheld NOISE 0.010 to 0.018 rad, 18 to 26 frame scalecamera rotationblockout skill; peak-to-peak measured 0.66 x strength

Camera language

ChoiceReads asSource
Eye level, medium lens, centered, staticneutral; the start for every A/BDillon, Hjalti
Low or tucked in a cornerhidden observerDillon
Wide lens from far, then a quick zoom inbeing watched; someone fixing on a targetDillon
Extreme wide lens close up"not okay", distortionDillon
Long lensinformational, flat, isolates, intensifies a two-character momentDillon, Hjalti, Spitfire
Top view; dolly zoom (linear keys)pressure from above; loss of controlDillon
Off-center from behindisolationDillon
Side-on movementthe most neutral screen motionDillon
Tilt or move ending on a detailthat detail becomes the subjectHjalti
Faceless framing, longer shots, significant object at the edge carried by soundroutineDillon

Moves that feel physical (Dillon): handheld noise when a watcher or a groggy body holds the camera, a camera operator's quick zoom, the camera dropping with the character's shoulders, less motion for laziness, more for urgency; a product shot is the stillest shot. Hjalti pulls focus across a handoff (dof.focus_object). Cuts: one side of the action line; an over-the-shoulder uses the shoulder the camera already occupies, and a jump cut can mean elided routine time (Dillon); a POV sits just off the line on the camera side [added, S6].

Quality gates

Measurable: P.audit(sc, table, chars, line=(a, b), travel={name: "right"}, stepped=movables): marker switch at each shot frame and not one frame before, main subject's eye line in frame and not blocked (ray cast names the blocker, fog included), measured vs intended size, height class with an intent for every non-eye camera, one side of the action line (POV and cross exempt), screen direction per journey, block length, two inserts in a row, non-constant keys outside live blocks, opaque passepartout, cost gates; warnings lists lights. Plus the edit runtime, ffprobe streams and duration, and for live shots an edit frame rendered against the previs frame.

Visual (open each image): contact sheet (every beat visible, one focal point per shot, hero readable, cheats and occluders invisible, pose readable from this camera), floor plan (camera side, journeys), A/B sheet (isolation vs context), frames pulled from the MP4 (order, burn-ins, push, live-shot content). Comic timing and sound stay a human call (Hjalti).

Common mistakes

MistakeLooks likeFix
Shots sprinkled over a finished performancegeneric coverage, wrong framingsdesign each shot from its beats (Hjalti)
Wide framing that ignores the extremesthe tallest pose or entrance clipsfit="all" and the extremes audit
POV from the true eye pointthe counter or the character's own head blocks the viewcheat the POV forward and up (S6: 0.6 m forward, 1.72 m high, over the counter); the audit reports the blocker
Door or prop between camera and herohero hidden in the key frameswing it away or move the camera; the contact sheet catches it
Over-the-shoulder from the far sideline crossed, hero out of framestay on the camera side; audit eye line
Unkeyed cheatearlier shot changeskey_block every movable at every block
6 fps board scene as a scene strip in a 24 fps editplays 4x fastexport stills or retime
Every shot animated, mood lighting, full rigsslow, the edit fights the moves, nothing judged soonerstills plus push, live only when the move is the meaning, Workbench grays, mannequins
Coverage as separate strips or moved keyscuts drift from the actioncover=True run, P.retime_cut
Fog world-aligned or in fronthero tinted, fog in every shotP.fog_for_shot (behind the subjects, one shot)

Blender 5.2 notes

  • bpy.ops.render.opengl fails in --background; use Workbench render.render (headless) or a camera-locked viewport in a live session.
  • keyframe_insert() ignores the new-key interpolation preference (keys stay BEZIER) though "Only Insert Available" is on: call P.make_stepped.
  • frame_set on a scene that is not the window scene leaves transforms stale, and a write_still render of such a scene uses the window scene's frame: P.activate, P.render_frame.
  • VSE: strips, left_handle/right_handle, content_start; SceneStrip.frame_start and frame_final_* warn (removal in 6.0). Strip.split(frame, split_method='SOFT') splits from data. New scenes default to AgX: set the edit scene to the source view transform.
  • marker.camera = cam binds headless; bpy.ops.marker.camera_bind needs a timeline context.
  • Rigify is off under --factory-startup: bpy.ops.object.armature_basic_human_metarig_add raises "could not be found" until P.ensure_rigify().
  • Selection Sets, Storypencil and Grease Pencil Tools are not bundled; use a bone collection layout. The Storyboarding app template ships an Edit scene with sync on and auto-key off in Edit (pin off in its Storyboarding workspace).
  • GP v3 strokes render in a headless Workbench still.

References

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Scenario Blender Previs Storyboard AI skill do?

Use when turning a script or story into previs, layout, storyboards or an animatic in Blender: beats to shot list, camera and lens per shot, staging, camera height, 180-degree rule, screen direction, shot sizes, when to cut, camera moves, timing stills in the Video Sequencer, markers bound to cameras, Story Tools, Grease Pencil or 2D/3D hybrid boards. Also when an animatic shows the wrong frame, blocking keys are not stepped, Rigify operators are missing, or viewport render fails headless.

Why use Scenario Blender Previs Storyboard on TypingMind?

Because you install it once and use it with any model. Scenario Blender Previs Storyboard is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Scenario Blender Previs Storyboard in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/scenario-labs/skills/tree/main/skills/dcc/blender/scenario-blender-previs-storyboard. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Scenario Blender Previs Storyboard?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Scenario Blender Previs Storyboard?

As many as you like. As long as a model supports skills, you can use Scenario Blender Previs Storyboard with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Scenario Blender Previs Storyboard AI skill free?

Yes. It is published on GitHub by scenario-labs under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇