Chapter 1: Introduction & Quick Start

1.1 What is Scribis?

Scribis is a next-generation AI speech-to-text, subtitle editor, and multimodal creation studio crafted for video creators, translators, researchers, podcasters, and professionals.

Unlike traditional cloud-based transcription tools that charge per minute and upload sensitive audio to remote servers, Scribis is built with a Local-First Privacy architecture. Your audio files and transcripts remain 100% on your device unless you explicitly connect your own cloud AI keys.

🌟 Key Product Highlights

1. 🔒 100% On-Device Privacy: Speech recognition and media rendering run directly on your GPU/CPU with zero cloud leakage.
2. ⚡ Top-Tier Open-Source ASR Models: Features world-class models including Parakeet v3 TDT (#1 for English with ultra-fast speed), VibeVoice (state-of-the-art context and zero hallucinations), SenseVoice Small (ultra-fast CPU inference with emotion detection), and Whisper Large-v3.
3. 🎬 Text-Driven Video Editing: Edit video by editing text. Jump playheads by clicking any word, perform rough cuts by deleting subtitle lines, and add dynamic zoom-in pans, privacy mosaics, and title overlays directly on the timeline.
4. 🎙️ Global Dictation & Mini Recorder: Press a global hotkey anywhere on your desktop to dictate smoothly with smart auto-formatting, grammar polishing, and instant paste into Word, Notion, IDEs, and messengers.
5. 🤖 AI Assistant & GEC Correction Sync: Chat with lengthy recordings, generate structured meeting summaries, and perform homophone error correction with one-click sync to your personal Correction Dictionary.

---

1.2 3-Step Quick Start Guide

Follow these three simple steps to transcribe your first audio or video file in under 3 minutes:

Step 1: Import Media Files

Open Scribis and click "Transcribe" in the left sidebar. Drag and drop your audio or video file (.mp3, .wav, .mp4, .mov, etc.) into the upload zone.

[!TIP] You can also click "Record from Mic" to record audio directly in the app.

---

Step 2: Choose ASR Model & Language

Configure your transcription settings in the right panel:

1. Model Selection:
   • For English Audio (#1 Recommended): Select Parakeet v3 TDT (fastest inference, lowest Word Error Rate).
   • For Maximum Precision & Multilingual Context (#1 Overall): Select VibeVoice.
   • For Low-Resource Laptops / CPU Inference: Select SenseVoice Small.
2. Language: Set to Auto or explicitly select your spoken language (e.g., English, Chinese (Traditional), Japanese).
3. Speaker Diarization: Turn on Speaker Diarization if your media contains multiple speakers.

Click "Start Transcription".

---

Step 3: Review, Edit & Export

Once transcription finishes, Scribis seamlessly opens the Translate & Subtitle Editor:

1. Interactive Review: Click any subtitle line to jump video playback to that exact timestamp. Edit words directly or use shortcuts (Alt + E edit, Alt + M merge, Alt + D delete).
2. AI Error Correction (GEC): Click "GEC Correction" to inspect suggested typo fixes in a red/green Diff Review panel and sync accepted fixes to your dictionary.
3. Export: Click "Export" in the top right corner to download .srt, .vtt, .txt, .fcpxml, or render a hardcoded burn-in MP4 video.