SuperAgent

1. SuperAgent Video Studio Update: Sora 2 Video Model Now Available

With this release, Video Studio now supports the following configurations when using Sora 2:

Aspect Ratios

16:9 — Horizontal (ideal for YouTube, websites, presentations)
9:16 — Vertical (ideal for TikTok, Reels, Shorts)

Video Duration

10 seconds
15 seconds

Quick Creator Workflow

When creating a video with Sora 2, you can now explicitly select:

Aspect Ratio
Duration (10s or 15s)
Number of videos to generate
Resolution

This gives you predictable outputs that align with your distribution channel from the very beginning.

2. SlideAgent Update: Manual Edit Mode for Precise, Hands-On Slide Editing

We’ve introduced Manual Edit mode in SlideAgent, giving you direct, canvas-level control over your slides after they’ve been generated by AI. This update is designed for moments when you want absolute precision—fine-tuning wording, adjusting emphasis, or making last-mile edits before sharing or presenting.

How it works

Enter Manual Edit mode
- Open an existing slide deck from the My Decks panel.
- Click Edit in the top toolbar and select Manual Edit from the dropdown menu.
Edit directly on the canvas
- Click any editable element (text blocks, titles, or content areas) to highlight and activate it.
- Make changes instantly - text updates are applied directly where you see them, without additional dialogs or prompts.
Real-time preview
- All changes are reflected immediately on the canvas, so you can evaluate layout, spacing, and visual balance as you edit.
Undo / Redo with confidence
- The bottom toolbar records every action in your current editing session.
- You can undo or redo steps freely, allowing you to experiment without risk.
Save and version your changes
- Click Save in the top-right toolbar to commit your edits.
- Each save creates a new version, ensuring your original AI-generated deck remains intact.

Why this matters

Greater control, zero friction: Make precise edits without regenerating slides or re-running prompts.
Perfect for final polish: Ideal for last-minute revisions before a client meeting, executive review, or presentation.
AI + human workflow: Start fast with AI, finish strong with manual refinement - on your terms.

With Manual Edit mode, SlideAgent now supports both AI-driven creation and hands-on professional editing, giving you the flexibility to move seamlessly between automation and control.

3. SlideAgent Update: Upload Images to Guide, Brand, and Supercharge Your Slides

You can now upload images directly when creating or refining a deck, enabling SlideAgent to see what you see, understand your intent more precisely, and generate slides that align with your real-world context.

Supported Image Upload Specifications

Supported formats: jpg, jpeg, png, webp
Max size per image: 5MB
Max number of images: 10
Total upload size limit: 10MB

How SlideAgent Understands Your Images

Once images are uploaded, SlideAgent’s LLM analyzes both your prompt and the visual content together.

Based on semantic alignment, SlideAgent automatically assigns one or more functional roles to each image.

This is not manual tagging — it’s intent-aware, context-driven reasoning.

Image Role Intelligence (Automatic Role Assignment)

Each uploaded image can serve one or multiple roles in the presentation pipeline:

1. THEME_STYLE_REF

Used as a visual reference.

Extracts dominant colors
Applies them as the global slide theme
Ensures visual consistency across the entire deck

Typical use: brand mood boards, inspiration screenshots, design references.

2. LOGO_BRANDING

Used as a brand identifier.

Automatically placed in master slides
Appears in headers, footers, or cover positions
Maintains consistent brand presence on every slide

Typical use: company logos, product marks.

3. CONTENT_ILLUSTRATION

Used as slide-level illustrations.

Matched to the most relevant slide by semantic meaning
Layout adapts dynamically based on aspect ratio
- Left image / right text
- Top image / bottom text
- Full-bleed visual blocks

Typical use: product screenshots, diagrams, conceptual visuals.

4. BACKGROUND_OVERLAY

Used as full-screen background imagery.

Automatically fills the slide background
Applies smart overlays to preserve text readability
Adjusts contrast, brightness, and mask intensity

Typical use: hero visuals, emotional or atmospheric imagery.

5. DATA_CONTENT_EXTRACT

Used as a knowledge source.

Extracts text, numbers, or chart insights from images
Integrates extracted data into slide content or outlines
Does not visually place the image unless relevant

Typical use: screenshots of reports, charts, dashboards, tables.

6. IGNORE

Used when an image is semantically irrelevant.

Image is excluded from layout and content
SlideAgent proceeds normally
User is informed that the asset was not used

Typical use: accidental uploads or unrelated visuals.

Typical User Scenarios

User Prompt	Uploaded Image	Assigned Role	Final Result
“Create a tech company intro deck”	FlashLabs blue logo	`THEME_STYLE_REF`, `LOGO_BRANDING`	Blue-and-white color scheme, with the logo placed in the top-right corner of each slide.
“Analyze H2 sales performance”	Screenshot of sales line chart	`DATA_CONTENT_EXTRACT`, `CONTENT_ILLUSTRATION`	Generate a sales analysis slide, with copy that incorporates the data shown in the image, and include the original image on the slide.
“Write a romantic travel guide”	Sunset beach photo	`BACKGROUND_OVERLAY`	Use the image as a full-screen background on the cover slide, with white text.
“Help me polish this slide”	Unrelated landscape photo	`IGNORE`	Optimize the text as usual, do not insert the image, and inform the user that the attachment was not used.

Others

4. SlideAgent Update: Unified Billing & Transparent Token Accounting

As SlideAgent evolves into a more powerful, multi-model AI system, clear and predictable billing becomes critical.

This ensures that every step of slide generation is accurately metered, traceable, and consistent with the rest of the FlashRev AI platform.

What You’ll See in Token Management

With this update, SlideAgent usage now appears clearly in Token Management:

Each model run is logged as a separate entry
- Feature category
- Model name
- Token consumption per execution
- Exact execution timestamp

FlashRev New Weekly Update [29th December 2025 - 31st December 2025]