Storyboards to Cinematic Video: First-Last Frame Pairing in Stensyl.

```html

Why First-Last Frame Pairing Solves the Motion Control Problem

Standard AI video generation treats each clip as a blank slate. The model interprets a text prompt, makes its own compositional decisions, and produces motion that may be technically impressive but compositionally unpredictable. Chain three of those clips together and you have three separate creative interpretations pretending to be one continuous sequence.

First-last frame pairing breaks that pattern by giving the model two fixed anchor points: the exact state of the scene at the start of the clip and the exact state at the end. Instead of inventing the path, the model calculates a motion path between two known positions. The creative decision space collapses from infinite to constrained, and that constraint is the feature.

The practical difference is significant. Freeform generation produces clips where lighting shifts between frames, subjects drift in and out of the intended composition, and camera moves reverse direction mid-shot. Paired-frame generation locks the anchor points, so the subject position, lighting direction, and framing remain consistent across the transition. What changes is only what you intend to change: the camera position, the depth of field pull, the progression through a space.

For film pre-visualisation, architectural walkthroughs, and exhibition flythrough sequences, this predictability is not a creative limitation. It is the professional baseline. Clients and directors need to see a coherent spatial sequence, not a highlight reel of interesting moments that happen to share a colour grade.

In a real production pipeline, this technique slots in after concept approval and before final render or client presentation. The storyboard frames have been signed off. The spatial logic of the sequence is agreed. First-last frame pairing translates that approved logic into motion without reopening compositional decisions that have already been resolved.

Setting Up Your Storyboard Frames in Stensyl

The workflow starts in Storyboards (/storyboards), Stensyl's scene-by-scene boarding surface. Before any video generation touches the timeline, the sequence needs to exist as a structured set of annotated frames.

Structuring Scenes and Annotating Frames

Build each scene as a discrete unit with a clear description of the opening composition and the intended closing composition. The annotation layer matters here. Camera direction notes — "slow push forward, left to right", "static wide, tilt up to ceiling" — carry meaning when you move to the Film studio. Lighting notes are equally important: "hard side light from east-facing window" or "overcast exterior, no shadow direction" give the generation model consistent reference rather than leaving it to sample from its training distribution.

Keep annotations precise rather than evocative. "Warm and cinematic" is mood direction for a director of photography on a live shoot. For AI generation, "3200K practical light source, top-right frame position, long shadow to lower left" is actionable. The model works with spatial and chromatic information, not emotional intent.

Generating Polished Reference Stills

Storyboard sketches are useful for sequencing logic, but they are rarely detailed enough to serve as video anchor frames directly. Use the Generate surface (/generate) to produce polished still images from storyboard descriptions before you pair them as video anchors. A rendered architectural interior at the correct camera angle and lighting condition is a far stronger anchor than a pencil thumbnail.

The practical step: take the frame description from Storyboards, pass it into Generate with any relevant reference images attached, and produce a still that matches your intended composition precisely. Adjust the output until the framing, lighting, and subject position are exactly right. That rendered still becomes the anchor frame.

Keeping Assets Linked with Projects

Use Projects (/projects) to keep storyboard assets, generated stills, and eventual video outputs linked under one workspace. In a production involving a three-scene architectural sequence, you will accumulate original sketches, multiple generated still variants, selected anchor frames, and rendered clips. Without a project structure, assets migrate across different tabs and output folders and the connection between a generated clip and the storyboard frame that produced it becomes unclear.

A project workspace also supports team-shared access, which matters when a production designer and a visual effects supervisor are working on the same sequence from different machines.

Before opening Film, export exactly two stills per scene: the intended opening composition and the intended closing composition. Label them clearly — "scene-01-first.jpg" and "scene-01-last.jpg" — and keep them inside the relevant Project folder. This single discipline removes ambiguity at every subsequent step.

Building the Video Sequence Inside Film

Film (/film) is Stensyl's multi-scene cinematic video studio. This is where anchor frames become a controlled sequence rather than a collection of isolated clips.

Uploading Anchor Frames and Configuring Each Scene

For each scene, the process follows three steps. First, upload the first-frame still you exported from Generate. Second, set the last frame using the corresponding closing still. Third, configure the scene duration and describe the motion intent — the type of movement, its speed, and any focal changes that should occur between the two anchor points.

The motion description between two locked frames is shorter and more specific than a freeform prompt. You are not describing a scene from scratch; you are describing a transition. "Slow forward dolly, slight upward tilt, no subject movement" tells the model exactly what the space between your two anchor images should contain.

Chaining Scenes for Full-Sequence Continuity

The technique that makes this approach production-viable is scene chaining. The last frame of scene one becomes the first frame of scene two. The model receives the same image twice: once as the output anchor of the previous clip and once as the input anchor of the next. This creates a visual handoff point that maintains subject position, lighting state, and spatial orientation across the cut.

Without this chaining discipline, even well-produced individual clips produce jarring transitions. The subject shifts position. The lighting direction changes. The camera appears to teleport. Chaining eliminates that discontinuity at the compositional level before any editing correction is needed.

Credit Cost and Concurrent Generation by Tier

Credit cost scales with scene count and duration. Longer scenes and higher output quality consume more credits per clip. Running multiple scenes simultaneously is constrained by your plan's concurrent generation allowance. The table below shows how that maps across Stensyl's four tiers.

Tier	Monthly Cost	Credits	Concurrent Generations
Lite	£10/mo	1,000	1
Starter	£22/mo	2,500	1
Pro	£42/mo	6,000	2
Studio	£84/mo	12,500	4

For a production-scale sequence — say, eight scenes for a client presentation — Studio tier's four concurrent generations means you can run two paired-frame groups simultaneously, halving the wall-clock time to a full sequence render.

A Practical Example: Three-Scene Architectural Walkthrough

Consider a three-scene walkthrough of a residential interior. Scene one moves from the entrance threshold to the living room entrance, anchored by a still of the front door at full aperture and a still of the living room archway. Scene two moves from the archway to the seating area, anchored by the archway frame and a still of the sofa arrangement in correct perspective. Scene three moves from the seating area to the feature window, anchored by the sofa still and a wide shot of the window wall.

Each scene transition is visually seamless because each clip shares an anchor frame with its neighbours. The client sees a single continuous spatial journey. The production team generated three discrete clips and chained them. The storyboard intent is preserved exactly.

Using Canvas to Script and Review Before You Generate

Generation credits are finite. Spending them on clips that fail because the motion brief was underspecified is the most avoidable cost in this workflow. Canvas (/canvas), Stensyl's node-based workflow editor, is where you eliminate that cost before it occurs.

Using the LLM Chat Node to Refine Shot Descriptions

The Canvas LLM Chat node gives you access to multiple writing models in a pipeline context. Pipe a raw scene description into the node and ask the model to sharpen it into a precise motion brief. The model picker inside Canvas offers six options across Stensyl's tiers. For this task, the practical choice depends on what you need: Gemini Flash is fast and sufficient for iterating on straightforward motion descriptions. Claude Sonnet 4.6 (available on Pro tier) produces more nuanced output for complex scenes where spatial relationships, atmospheric conditions, and subject behaviour need to be described with precision.

A typical scripting pass takes ten to fifteen minutes per scene. You enter the storyboard description, ask the LLM Chat node to produce a paired-frame motion brief, review the output, and adjust. The result is a motion description that is specific enough to constrain the model's output without being so rigid that it cannot interpret the transition naturally.

Validating Stills Inside Canvas Before Film

Canvas also supports an image Generate node, which you can wire into the scripting sequence to validate a still before promoting it to a Film anchor frame. If the Generate output inside Canvas confirms that the composition matches the storyboard intent, the same prompt parameters carry directly into the Film setup. If it doesn't, you iterate inside Canvas at lower cost than re-running a full video generation.

This is not a redundant step. A still that looks correct as a standalone image sometimes fails as a video anchor because the framing doesn't give the model enough spatial information to interpolate motion. Catching that in Canvas avoids discovering it after a video generation credit has been spent.

Using Ray to Sanity-Check Model Choices

Ray (/ray) is Stensyl's creative-decision assistant. Before committing to either an image-first path (Generate, then Film) or direct video generation inside Film, Ray can help clarify which approach suits a given scene. Describe the scene and the type of motion you intend, and Ray will recommend whether the spatial complexity warrants producing a high-quality still anchor first or whether the scene is simple enough to approach directly.

Ray runs on a fast Anthropic Haiku model, which keeps response time low. It is the right tool for quick directional questions, not extended creative development. Use it as a checkpoint, not a collaborator.

The Canvas scripting pass costs no generation credits. Every scene description refined in Canvas before Film is a credit protected rather than a credit spent on a recoverable mistake.

Editing and Finalising the Output

A rendered sequence from Film is strong but rarely finished. Colour grading may need adjustment to match a client's brand palette. A rendered object in one scene may need clean-up. Compositing an overlay — a logo, a title card, a before-and-after split — requires frame-level access.

Frame-Level Corrections in Editing

The Editing surface (/editing) handles this layer. It is desktop only, which means it is not available in a browser session on a mobile device, but for production work that is rarely a constraint. Bring individual frames or clips from the Film render into Editing for colour grading, object clean-up, or compositing adjustments. The surface operates at the frame level, so corrections that would take significant time in a separate application can be handled inside the same platform where the sequence was generated.

Keeping the Output Aligned with Moodboards

During the editing pass, keep Moodboards (/moodboards) open as a reference layer. The moodboard assembled at the start of the project captured the colour temperature, material palette, and spatial atmosphere the sequence was intended to express. Editing decisions made without that reference can drift. A colour grade that looks technically correct in isolation may shift the mood away from what was approved at the concept stage.

The discipline here is simple: make an edit, compare against the moodboard, confirm the output still belongs to the same visual world. It adds minutes to the editing pass and prevents hours of revision after a client presentation.

Exporting for Delivery and Post-Production Handoff

Export the finished sequence for client delivery, presentation decks, or handoff to a post-production team. For presentation contexts, a compressed MP4 is standard. For post-production handoff, confirm the export format against the receiving team's pipeline requirements before rendering.

Repurposing Clips via Social

Short clips from the sequence — a single dramatic scene from an architectural walkthrough, a product reveal moment — can be repurposed directly via the Social surface (/social-studio). Carousel formats, short-form video posts, and reel-style outputs can be formatted without leaving Stensyl. For studios that deliver both a client presentation and social content from the same project, this eliminates a separate repurposing pass in a different tool.

Final Quality Check

Before marking the sequence as delivered, compare the exported video frame by frame against the original storyboard. This is not a creative review. It is a technical confirmation that the paired-frame technique held visual intent across every scene transition. Check that subject position at each cut point matches the storyboard's intended composition. Check that lighting direction is consistent across the full sequence. Check that the motion path within each scene corresponds to the annotated camera direction.

If a scene fails this check, the storyboard is the reference, not the render. Identify which anchor frame diverged from the approved composition and regenerate that scene with a corrected still.

Where This Workflow Fits Each Design Discipline

The paired-frame technique is not a specialist tool for one type of project. The underlying logic — constrain the model between two known states, chain scenes to maintain continuity — applies wherever a client needs to experience a designed space or object in motion before it physically exists.

Architecture and Interior Design

Use first-last pairing to simulate a camera dolly through a rendered space, moving from entrance to focal point. An architectural visualisation that shows the progression from a building's lobby to its atrium communicates spatial scale and materiality in a way that a static render cannot. The technique produces that sequence without a 3D animation pipeline and without a survey of physical location.

Game Design and Exhibition Design

Chain scenes to produce a flythrough of a game environment or a visitor journey through an exhibition installation before any 3D build begins. For exhibition designers, this is particularly valuable at the pitch stage: a credible spatial sequence generated from storyboard frames can secure a commission before a single physical element is specified. The paired-frame approach keeps the sequence spatially coherent even across scenes that represent significant changes in environment scale.

Film and Set Design

Generate pre-visualisation sequences from rough storyboards to share with directors before committing to physical set dressing. A pre-vis sequence produced via Film from annotated storyboard frames communicates camera logic, spatial blocking, and lighting intent without the cost of a full pre-production build. Directors can respond to a moving sequence in ways they cannot respond to a static frame, and the feedback gathered at that stage is far cheaper to act on than feedback received after the set is dressed.

Product and Automotive Design

Anchor frames on a product's front three-quarter view and rear three-quarter view to generate a controlled reveal clip. For an automotive design presentation, this approach produces a 360-style reveal without a physical model or a studio shoot. The constraint of two anchor frames keeps the model's interpretation of the product surface consistent. Freeform video generation of a vehicle tends to drift on reflective surface detail and silhouette precision. Paired frames hold both.

The Cross-Discipline Benefit

Across all nine design disciplines Stensyl serves, the paired-frame approach reduces the number of generation iterations required to reach a usable output. The model has clear compositional boundaries on both ends of the clip. It is not making creative decisions about subject placement or camera position. Those decisions were made at the storyboard stage, confirmed in Generate, and locked into Film as anchor frames.

The result is a workflow where creative intention set at the beginning of a project survives through to the delivered sequence, rather than being gradually reinterpreted by successive generation passes. That survival of intent is what professional pre-visualisation requires, and it is what the first-last frame pairing technique, executed correctly across Stensyl's surfaces, consistently delivers.

```