ScribeShot
Notes & summaries
Turn transcript material into structured notes and concise summaries.
Your AI. Your Mac.
ScribeShot runs supported AI inference locally on your Mac. There is no AI API key to configure and no hosted ScribeShot AI backend in the current product. Choose a smaller Gemma model for efficiency or Qwen3-VL when higher-quality vision, OCR and reasoning matter more.
Product docs: Set up ScribeShot with local Gemma AI →
Local AI powers the learning workflow
Local AI is not a side feature. It powers the parts of ScribeShot that turn a YouTube video into material you can understand, retrieve and practice later.
Chapters
Break long videos into useful, timestamped sections.
Screenshot understanding
Use captured slides, code, diagrams and visible text as part of the learning context.
Chat & knowledge
Ask about one video or reuse saved knowledge across your library.
Learning & quizzes
Create study material and active-recall questions from what you saved.
Mind maps
Turn a video's ideas into an interactive visual hierarchy.
Choose speed or quality — both stay local
ScribeShot keeps the model choice simple: pick the footprint and quality tier that fits your Mac and the kind of material you process most often.
| Model | Approx. download | Best fit |
|---|---|---|
| Gemma 4 E2B | ~2.0 GB including vision support | Lightest supported option; a good starting point for an 8 GB Mac. |
| Gemma 4 E4B (default) | ~3.6 GB including vision support | Best everyday balance of speed and output quality. |
| Qwen3-VL 8B | ~5.8 GB including vision support | Higher-quality local vision, OCR and reasoning; recommended for 16 GB+ systems. |
What “local” means
After you download a supported model, ScribeShot performs supported AI generation and analysis on your Mac rather than sending that AI workload to a hosted ScribeShot AI service.
ScribeShot is still source-connected. Internet access is required to open YouTube, fetch available transcripts and metadata, download models, and check for app updates. Local AI describes the AI-processing boundary, not a promise that every part of the product works with the network disconnected.
Why local AI matters
- No AI API key: there is no provider account or key to configure for the current AI workflow.
- Private processing boundary: supported AI work runs on your Mac.
- Predictable cost: local inference does not create a per-request AI bill.
- Model choice: use a lighter model for efficiency or a larger local tier when quality matters more.
- Files you control: notes, transcripts, screenshots and generated learning artifacts remain ordinary files in your chosen library.
Start with the default
Gemma 4 E4B is the default for everyday use. Choose E2B when memory and speed matter most, or Qwen3-VL 8B on a 16 GB+ Mac when you want the higher-quality local vision and reasoning tier.
Common questions
Do I need an AI API key?
No. The current ScribeShot AI workflow uses supported local models.
Which model should I pick first?
Gemma 4 E4B is the default. Use E2B for the lightest footprint or Qwen3-VL 8B for the higher-quality local vision and reasoning tier.
Can I use ScribeShot completely offline?
Supported AI inference can run locally after the model is downloaded, but YouTube source access, model downloads and updates still require internet access.
Run it offline on your own Mac
The local model is included, not an add-on. Download the app, pick a Gemma 4 model, and generate notes with no key and no account.
Try 3 videos free • 3 screenshots per video • No credit card • Zero AI cost with local LLM