Video Storyboard & Approval

The storyboard is the pre-production review step between script generation and final video rendering. You see every scene laid out visually -- image, voiceover text, timing, mood -- and decide what stays, what gets regenerated, and what gets edited before committing to the full render.

What It Does

When you create a video in ILLIXIS, the AI first generates a script with scene breakdowns, narration text, visual descriptions, and music direction. The storyboard takes that script and produces the actual assets you will hear and see in the final video: scene images, a voiceover audio track, and background music.

The storyboard exists so you never produce a video blind. You review each element, approve what works, regenerate what does not, and edit narration text directly. Only after you approve every scene does the "Generate Video" button unlock. This prevents wasted render time and cost on a video you would have rejected.

The storyboard is audio-first. Voiceover is generated from your narration text using AI voice generation, and each scene's video clip duration adapts to the actual spoken length of that scene's narration. This means what you hear in the storyboard preview is exactly what plays in the final video -- no speed manipulation, no compression artifacts.

The Storyboard Workflow

The full pipeline from idea to finished video follows this sequence:

Steps 3 and 5 are where you spend your time. Everything else is automated.

Storyboard Layout

Once storyboard assets finish generating, the page shows three sections:

Script & Narration

The top card displays your full narration text organized by scene. Each scene shows:

• Scene number and segment label (Hook, Problem, Solution, Proof, CTA, etc.)
• Narration text -- the exact words the voiceover reads
• Visual description -- what the scene image depicts
• Duration -- actual spoken duration from voiceover word timestamps, rounded up to whole seconds

A duration bar at the top shows the total video duration compared to your target (e.g., "Video: 28s / 30s target"). This is the sum of per-scene clip durations, which is what ad platforms measure.

Audio Controls

Two audio players let you preview:

• Voiceover -- the full narration track with playback controls
• Background Music -- the AI-generated music track

Each has a regenerate button with remaining count displayed.

Scene Cards

A grid of cards, one per scene. Each card shows:

Reviewing Scenes

Work through the storyboard from top to bottom:

1. Check the duration bar. If the total video duration exceeds your target, the bar turns amber and tells you how many seconds to trim. You fix this by shortening narration text.
2. Listen to the voiceover. Play the audio to hear pacing and pronunciation. If something sounds off, you can edit the narration and regenerate.
3. Listen to the music. Make sure the mood fits. Regenerate if it does not match.
4. Review each scene card. Check that the image matches the visual description, the narration makes sense for the scene, and the duration is reasonable.
5. Approve or regenerate. Click "Approve" on scenes that work. Click "Regen" on scene images that need a new version.

Editing Voiceover Text

You can edit narration text at two stages:

Before Storyboard Generation (Script Review)

On the script review screen (before clicking "Generate Storyboard"):

1. Click Edit in the script card header.
2. All scene narration fields become editable text areas.
3. An estimated voiceover duration bar appears, updating in real time as you type.
4. Edit any scene's narration, visual description, text overlay, or mood.
5. Click Save Changes.
6. Then click Generate Storyboard to produce assets from your edited script.

After Storyboard Generation

Once the storyboard exists with voiceover and images:

1. Click Edit Script in the Script & Narration card.
2. Narration fields for each scene become editable.
3. A live estimate shows your new projected duration vs. the current voiceover duration.
4. Edit the narration text.
5. Click Save & Regenerate Voiceover.

This saves your text changes and immediately triggers a new voiceover generation. The page polls automatically and reloads when the new voiceover is ready. Scene durations update to reflect the new spoken timing.

This is the primary way to fix duration compliance issues. If your video is 5 seconds over the target, shorten the narration across a few scenes, then Save & Regenerate.

Speed Control

The voiceover speed slider sits next to the voiceover regenerate button. It controls the speaking rate of the AI voice.

When your video duration exceeds the target, the storyboard suggests a speed value. For example, if your voiceover is 35 seconds but your target is 30 seconds, you will see "Suggest: 1.17x" as a clickable link that pre-fills the speed input.

To use:

1. Set the speed value in the input field (or click the suggested speed link).
2. Click the voiceover Regen button.
3. Confirm the regeneration (costs ~\/bin/zsh.02).
4. The page reloads with the new voiceover at the adjusted speed.

Editing narration text is generally more effective than increasing speed. Faster speech can sound unnatural. The recommended approach is to trim a few words from the longest scenes first, then use speed as a fine-tuning tool for the last 1-2 seconds.

Regenerating Elements

Each element type has its own regeneration limit per video:

Scene Image Regeneration

• Click Regen on any scene card.
• The image fades to 50% opacity while generating.
• A new image is produced from the same visual description and segment composition rules.
• The page reloads when the new image is ready.
• The regeneration count is shared across all scenes. With 6 scenes, you get 6 total regenerations to distribute however you want (e.g., 3 on Scene 1, 2 on Scene 4, 1 on Scene 6).

Voiceover Regeneration

• Click the voiceover Regen button in the Audio section.
• Optionally set a different speaking speed first.
• The new voiceover generates from the current narration text across all scenes.
• Per-scene durations update based on fresh word timestamps.

Music Regeneration

• Click the music Regen button in the Audio section.
• A new track is generated matching the script's music direction (style, energy arc, tempo).

When remaining regenerations drop to 2 or fewer, the button turns amber as a visual warning. At 0 remaining, the button is disabled.

Duration Compliance

Ad platforms reject videos that exceed standard duration units. The storyboard enforces duration compliance visually:

Target durations are set when you create the video (15s, 30s, 45s, 60s, or custom).

The duration bar at the top of the Script & Narration section shows:

The displayed duration is the video duration (sum of per-scene clip durations), not the raw voiceover duration. Video clips round each scene's spoken duration up to the nearest whole second, so the video is always slightly longer than the voiceover.

Per-scene durations on each scene card show the actual clip duration with color coding:

• Green: at or under the scene's target duration
• Amber: over the scene's target duration

To fix an over-target video:

1. Identify the longest scenes in the Script & Narration section.
2. Click Edit Script.
3. Trim a few words from those scenes' narration.
4. Watch the real-time estimate update.
5. Click Save & Regenerate Voiceover.
6. Alternatively, increase voiceover speed using the speed slider.

Scene Images

Scene images are AI-generated at 2K resolution in 9:16 (vertical) aspect ratio. Key details:

• Positive-framing prompts. The image generation system describes what each shot IS rather than what it is not. This produces cleaner, more unified compositions and reduces split-screen or composite artifacts.
• Segment-aware composition. Each scene's image follows composition rules based on its ad segment (Hook, Problem, Solution, Proof, CTA). Hook scenes use dramatic close-ups. CTA scenes use minimal backgrounds for logo overlay. Solution scenes include the product.
• Reference image consistency. The first scene with a person becomes the character reference for all subsequent people scenes, maintaining visual consistency across the video.
• Product references. Solution and proof scenes use your product images as references so the AI renders the actual product, not a generic substitute.
• 2K for storyboard, 4K for final render. Storyboard images generate at 2K resolution for faster iteration. The final video render upgrades to 4K resolution.

Audio Playback

Both audio elements have standard audio players in the storyboard:

• Voiceover player plays the complete narration track. Use this to check pronunciation, pacing, and natural flow.
• Music player plays the full background track. In the final video, music plays at reduced volume behind the voiceover.

Audio URLs include cache-busting parameters, so after regeneration the player always loads the fresh version.

Approving and Starting Production

Once you have reviewed and are satisfied with all elements:

1. Approve each scene by clicking the "Approve" button on each scene card. The badge changes from "Pending" to "Approved."
2. Or click Approve All in the cost summary bar at the bottom to approve every scene at once.
3. The Generate Video button enables only when all scenes are approved.
4. Click Generate Video to start full production (Phases 2-4: video clips, composition, multi-aspect export).
5. A cost confirmation appears. Confirm to proceed.
6. The page polls automatically and redirects to the video output page when production completes.

You can unapprove a scene at any time before starting production by clicking "Unapprove." This re-disables the Generate Video button until you approve it again.

The cost summary at the bottom shows:

• Pre-production cost -- what you have already spent on storyboard assets (images + voiceover + music)
• Production cost -- estimated cost for video clip generation and composition
• Total estimated cost -- combined pre-production and production

Related Features

• Video Production -- Full overview of the video production dashboard and pipeline
• Video Scripts -- How script generation works, ad type templates, scene breakdown structure
• Voice Library -- Managing voices for voiceover, custom voice cloning
• AI Video Generation -- Video model options, duration logic, motion prompts