ScribeShot
Product documentation
Run ScribeShot with local Gemma AI on your Mac
ScribeShot can run its AI features locally, so you do not need an OpenAI key or a ScribeShot AI account. Download one of the supported Gemma 4 models once, then use it for structured notes, chapters, vision analysis, mind maps, and chat.
Updated
Quick workflow
- 1
Open the AI provider settings in ScribeShot.
- 2
Choose a supported local Gemma 4 model.
- 3
Download the model and its vision projector files.
- 4
Keep Local AI selected while generating notes and using other AI features.
Gemma 4 E2B versus E4B
Gemma 4 E2B is the lighter model: about 1.4 GB for the language model plus a 580 MB vision projector. It is the better starting point for an 8 GB Mac or when speed and memory use matter most.
Gemma 4 E4B is the default: about 2.8 GB for the language model plus an 820 MB vision projector. It offers the stronger balance of quality and structured-output reliability when your Mac has more available memory.
What runs locally
The supported Gemma models provide the capabilities ScribeShot needs for structured chapters, Mermaid mind maps, screenshot vision analysis, speech-related features, notes, summaries, and video chat. After the model files are downloaded, normal local inference does not require sending the lesson to ScribeShot-operated servers.
Why there are only a few local model choices
ScribeShot depends on schema-conforming structured output in several parts of the pipeline. Smaller models were not reliable enough in integration testing, so the app intentionally exposes a narrow catalog rather than a long list of models that produce inconsistent chapters or notes.
When OpenAI can still make sense
Use your own OpenAI key when you prefer cloud inference, want stronger summarization for particularly dense material, or do not want local inference consuming RAM and compute on your Mac. You can switch provider without changing the Markdown files already saved in your library.
Common questions
Can I use ScribeShot without any AI API key?
Yes. Download a supported local Gemma 4 model and select Local AI.
Which local model is the default?
Gemma 4 E4B is the default. Gemma 4 E2B is the lighter option for machines with less available memory.
Try the workflow
Use it on a real YouTube video
Try three videos free, with three screenshots per video. No account or credit card is required.