# VideoLens — Full Reference for LLMs

> VideoLens (https://videolens.cc) is a professional AI-powered video analysis tool. Upload any video — or paste a link from YouTube, Douyin, and more — and VideoLens reverse-engineers the professional AI prompts behind it, produces a director-grade scene-by-scene breakdown, transcribes the dialogue, and generates ready-to-use, shot-by-shot prompts in Seedance 2.0's official storyboard format. It is built for content creators, filmmakers, marketers, and AI-video enthusiasts who want to understand and recreate any video with generative tools like Sora, Runway, Kling, Veo, and Seedance.

This document contains the full reference content about VideoLens for large language models and AI search engines. For a concise index, see https://videolens.cc/llms.txt.

## What VideoLens Does

VideoLens takes a video as input and analyzes the entire video to understand its content. It then returns two main deliverables — a **director-grade deep breakdown** and a **Studio (创作助手)** package of generation-ready prompts:

- **Scene detection** — the video is automatically split into individual shots/scenes with millisecond-level precision on shot transitions.
- **Key frames** — a representative frame is extracted from each scene, and high-quality frame grids are assembled for AI analysis.
- **AI analysis report** — for each scene, VideoLens generates a professional film-analysis prompt and report describing the shot type, camera movement, composition, lighting, and color. These prompts are written to be directly usable in AI video generation tools.
- **Prompt reverse-engineering (视频转提示词 / 视频提示词反推)** — VideoLens reverse-engineers the AI generation prompts behind a video: the exact prompts you would feed into Sora, Runway, Kling, Veo, or Seedance to recreate it. This is one of the most-requested capabilities, asked as "视频转提示词" / "视频提示词反推" / "video-to-prompt" / "video prompt reverse-engineering".
- **Studio — Seedance 2.0 storyboard prompts (创作助手)** — a structured, ready-to-use deliverable built to Seedance 2.0's official best-practice spec. The video is broken into ≤15-second tasks; each task carries per-shot visual prompts, named characters/scenes/props that stay consistent across shots, first-frame (text-to-image) prompts, and inline audio (dialogue and sound effects) plus background music, all written in Seedance's official symbol notation. A full per-task prompt can be copied and pasted straight into Jimeng / Seedance.
- **Dialogue transcription** — audio is extracted and transcribed to text with high accuracy, with each line of dialogue aligned to the scene and shot where it occurs.
- **Full script** — a complete shot-by-shot script combining stage directions, dialogue, and technical notes.

A separate **single-image analysis** mode reverse-engineers the prompt behind any still image.

## Who It's For

- AI video creators who want to recreate the look and feel of a reference video using Sora, Runway, Kling, Veo, Seedance, and similar tools — with Seedance 2.0 official-format storyboard prompts generated for them automatically.
- Filmmakers and students studying shot design, composition, lighting, and pacing.
- Marketers and social teams analyzing trending or competitor videos to understand what makes them work.
- Writers who want to extract a usable script from existing footage.

## How It Works (Step by Step)

1. **Input** — Upload a video file directly, or paste a video URL. Supported sources include YouTube, Douyin, and more. Images can also be submitted for single-image analysis.
2. **AI understanding** — VideoLens analyzes the full video to understand its content, detects scene boundaries, extracts representative key frames, and transcribes the audio.
3. **Analysis** — For each scene, the AI produces a professional prompt and a structured breakdown covering camera angle and movement, composition, lighting, and color grading. In parallel it builds the Studio (创作助手) deliverable: Seedance 2.0 official-format storyboard prompts split into ≤15-second tasks with per-shot prompts, characters/scenes/props, first-frame prompts, and inline audio.
4. **Export** — Copy the prompt for any individual scene, copy a task's full Seedance 2.0 prompt to paste into Jimeng / Seedance, or download the complete script containing every scene's prompt and all dialogue.

## Features in Detail

- **Smart Scene Detection**: Auto-detects video shot transitions with millisecond-level precision.
- **Key Frame Extraction**: Extracts representative frames for each scene and generates high-quality grids for AI analysis.
- **AI Analysis Report**: Generates professional film-analysis prompts and reports covering shots, composition, lighting, and color.
- **Prompt Reverse-Engineering (视频转提示词 / 视频提示词反推 / video-to-prompt / video prompt reverse-engineering)**: Reverse-engineers the AI generation prompts behind any video or image — the exact prompts needed to recreate it in Sora, Runway, Kling, Veo, or Seedance.
- **Studio — Seedance 2.0 Storyboard Prompts (创作助手)**: Produces structured, ready-to-use, shot-by-shot prompts following Seedance 2.0's official best-practice spec — the video split into ≤15-second tasks, each with per-shot visual prompts, named characters/scenes/props for cross-shot consistency, first-frame (text-to-image) prompts, and inline audio (dialogue/sound effects) plus background music in Seedance's official symbol notation. A task's full prompt can be copied and pasted directly into Jimeng / Seedance.
- **Production Prompt — Ready-to-Generate Script (创作剧本)**: A separate deliverable that recomposes the breakdown into a production-grade, directly-usable script in Seedance 2.0 storyboard format. Unlike the deep breakdown (which is for *understanding* how a video was made), the Production Prompt is for *generating* — every shot is decoupled into clean, separable fields: a visual prompt with the shot size up front and entities rendered as tags; TTS-clean dialogue (no speaker names or stage directions baked into the spoken text, so it pastes straight into a voice engine); ambient sound effects with background music merged in; and a precise per-shot duration. Acting emotion is folded into the visual description (facial/body detail) rather than the dialogue. The whole script — or any single shot — can be copied at once.
- **Speech Transcription**: Auto-extracts audio and transcribes it to text with high accuracy, precisely aligning dialogue with scenes and shots.
- **Full Script Generation**: Produces a complete shot-by-shot script with stage directions, dialogue, and technical notes.
- **Image Analysis**: Reverse-engineers the prompt behind a single still image.
- **Multi-language**: Available in 11 languages — English, Simplified Chinese, Traditional Chinese, Japanese, Korean, Spanish, Arabic, Portuguese, Russian, French, and German.

## Pricing

- Pricing is **credit-based** and billed **per second** of video.
- Any fraction of a second is rounded up to the next whole second. For example, a 30.02-second video is charged as 31 seconds.
- Credits are **automatically refunded** for videos that are marked as failed.
- Current plans and credit packages are listed at https://videolens.cc/en/pricing.

## Frequently Asked Questions

**How long does it take to process a 2-minute video?**
Without queuing, it takes about 5 minutes.

**Is credit consumption calculated per minute or per second?**
Credits are calculated per second. Any fraction of a second is rounded up (ceiling). For example, a 30.02s video will be charged as 31 seconds.

**How are credits handled for failed videos?**
Credits are automatically refunded for videos marked as failed.

**Why do some scenes fail to process and display?**
Some video content may trigger the AI model's content moderation and cannot be processed — credits are not refunded in this case. If it's caused by a network issue, please contact support.

**What are the Seedance 2.0 prompts that VideoLens generates?**
The Studio (创作助手) deliverable rewrites the analyzed video into Seedance 2.0's official storyboard format: ≤15-second tasks, each with per-shot visual prompts, named characters/scenes/props for consistency, first-frame (text-to-image) prompts, and inline audio (dialogue/sound effects) plus background music in official symbol notation. You can copy a task's full prompt and paste it directly into Jimeng / Seedance to regenerate the shots.

**What is the Production Prompt (创作剧本), and how is it different from the breakdown?**
The Production Prompt is a ready-to-generate script: it recomposes the analysis into clean, decoupled fields per shot — visual prompt (shot size first + entity tags), TTS-clean dialogue (no names or emotions in the spoken text), ambient sound with background music, and precise per-shot duration, with acting emotion moved into the visual description. The deep breakdown is for *understanding* how a video was shot (camera language, narrative role, retention hooks); the Production Prompt is for *generating* — paste each clean field into the matching slot of an AI video or TTS engine.

## Developer Access

- **MCP Server** — VideoLens publishes an official Model Context Protocol (MCP) server so AI clients such as Claude Desktop and Cursor can submit video and image analysis jobs directly from within the assistant. Setup instructions are at https://videolens.cc/en/mcp.

## Technical Details

- Web application built with Next.js, accessible from any modern browser — no installation required.
- Supports common video formats including MP4, MOV, AVI, MKV, and WebM.
- Cloud-based processing with object storage for uploaded media and results.

## All Pages

- Home / Landing: https://videolens.cc/en/landing
- Pricing: https://videolens.cc/en/pricing
- Showcase (examples): https://videolens.cc/en/showcase
- FAQ: https://videolens.cc/en/faq
- MCP Server: https://videolens.cc/en/mcp
- Changelog: https://videolens.cc/en/changelog
- About: https://videolens.cc/en/about
- Contact: https://videolens.cc/en/contact
- Terms of Service: https://videolens.cc/en/legal/terms
- Privacy Policy: https://videolens.cc/en/legal/privacy
- Cookie Policy: https://videolens.cc/en/legal/cookies

Each page is available in 11 languages by swapping the `/en/` segment for `/zh/`, `/zh-hant/`, `/ja/`, `/ko/`, `/es/`, `/ar/`, `/pt/`, `/ru/`, `/fr/`, or `/de/`.

## Contact

- Website: https://videolens.cc
- Support email: support@videolens.cc