What the Tool Does NOT Do
Explicit boundaries, sandbox limits, and platform technical clarifications.
To ensure absolute transparency and prevent unrealistic expectations, this page outlines the explicit technical boundaries of Serelius Flow Video.
1. Platform and AI Model Boundaries
Understanding the tool's runtime architecture is essential:
- Not a Proprietary AI Model: The tool is an orchestrator and video assembler. All images, videos, and texts are generated by Google Flow models under Flow's standard credit policies and safety filters.
- Does Not Run Outside Google Flow: Operates strictly as a Flow Tool within Google Flow. There is no standalone server or external SaaS account.
- No Perfect Visual Consistency Guarantee: While 1:1 character sheets and visual references significantly anchor character likeness, subtle variations depend on the generative AI model.
- No Video File Uploads: Accepts only still reference images (PNG, JPEG, WebP) for character sheets and props; video file uploads are not supported.
2. Rendering and Export Clarifications
Export rules adhere to modular production standards:
| Boundary | Concrete Behavior | Recommended Workflow |
|---|---|---|
| Single Merged Video | Does not render a single concatenated master file. | Downloads individual numbered MP4 clips plus a CMX 3600 EDL for NLE assembly. |
| Subtitle Extensions | Plain text assets download as .txt files. | Rename .txt to .srt, .vtt, or .edl after download. |
| Text on Images | AI models are actively instructed not to draw text. | Subtitles and captions are rendered cleanly by client-side code during Phase 5. |
| Fixed Style Presets | No hardcoded catalog of 100+ render styles. | Provides one validated default style; users write custom styles or request AI suggestions. |
| In-Flight Cancellations | Cannot cancel a generation once accepted by Flow. | Use PAUSE QUEUE to stop subsequent generations from dispatching. |
3. Audio, Subtitles, and Framing Limits
Audio handling and scene duration operate within strict boundaries:
- No Master Audio Ingestion: Does not align external audio tracks, detect speech silence, or trim video to uploaded MP3s. Audio is generated directly by Google Flow.
- Scene Durations: Fixed to 8 seconds per scene on Veo 3.1 models; Omni 1.1 Flash accommodates 4, 6, 8, or 10-second scenes.
- No Per-Scene Framing Selector: Framing is governed by script tags and the global project default.
- No Vector HUD Overlays: Deprecated
COMPONENTEtags are recognized and ignored with a warning; use Phase 2ASSETSfor graphics.
4. Operational and Sandbox Boundaries
Security constraints within the Google Flow iframe environment:
- No Popups or Direct External Links: Due to iframe sandbox policies, browser popups and
window.opencalls are disabled; all downloads route throughFlow.download. - Interface vs Content Translation: Changing interface language translates UI labels and errors only; it does not translate the user's script or spoken audio.
- No Built-in DRM or Watermarks: The current tool includes no software trial locks, watermark overlays, or account billing checks.
Nuxfire Production Kit
Ready to build and launch your SaaS?
Get 100% full source code ownership, no proprietary wrappers, with architecture ready for millions of requests on Cloudflare.