Scriberr Tutorial — Set Up Offline AI Transcription in 10 Minutes

📖 How to Set Up Scriberr: From Zero to First Transcription in 10 Minutes

I'll walk you through getting Scriberr running with Docker, uploading your first audio file, and using the AI chat feature to summarize a meeting. This should take about 10 minutes, maybe 15 if it's your first time running a Docker container.

Prerequisites

• Docker installed (no GPU required, but it helps)
• At least 4GB free RAM — models need memory
• About 2GB free disk for the model download on first run
• An audio file to transcribe (MP3, WAV, M4A, or video formats work)

Step 1: Start Scriberr with Docker

I chose port 8080 here so it doesn't conflict with anything else running on my server. Pick whatever works for you.

mkdir -p ~/scriberr/data
cd ~/scriberr

docker run -d \
  --name scriberr \
  -p 8080:8080 \
  -v $(pwd)/data:/app/data \
  -v $(pwd)/whisperx-env:/app/whisperx-env \
  --restart unless-stopped \
  testype/scriberr:latest

If you see a container ID printed — congratulations, it's running. Check with docker logs scriberr — you should see it starting up and downloading the model. This takes about 1-2 minutes on the first run. Grab a coffee.

Step 2: Open the Web UI

Open your browser and navigate to http://localhost:8080. You should see the Scriberr homepage:

Create an account (local only — no email verification needed, it's completely offline). Then you're in.

Step 3: Upload and Transcribe

Click the "Upload" button or drag-and-drop an audio file. I'm using a 20-minute meeting recording (MP3, 32MB).

Scriberr processes it locally — you can watch the progress bar. On my machine (4 vCPU, 8GB RAM, no GPU) a 20-minute file took about 3 minutes with the Parakeet model. That's impressive for fully local processing.

The transcript view shows word-level timing. You can click on any word and jump to that point in the audio. Speaker labels appear on the left — green for Speaker 1, blue for Speaker 2, etc.

Step 4: Chat With Your Transcript (The Best Part)

This is where Scriberr shines. Click the "Chat" icon next to your transcript. You'll need to configure an LLM backend first:

1. Go to Settings → AI Provider
2. Select Ollama (or OpenAI if you prefer cloud)
3. If using Ollama, enter the URL: http://host.docker.internal:11434
4. Pick a model — I use llama3.1:8b, works great for summarization

Now try these prompts:

• "Summarize this recording in 3 bullet points"
• "What were the key decisions made?"
• "Extract all action items with who they were assigned to"
• "Was there any discussion about the budget?"

The AI answers based on your transcript — it's like having a personal assistant who actually listened to the whole meeting and remembers everything.

Step 5: Automate With Folder Watcher (Optional)

Scriberr can monitor a folder and automatically transcribe any new file dropped into it. To set it up:

1. Create a directory on your host: mkdir -p ~/scriberr/watch
2. Mount it into the container: add -v ~/scriberr/watch:/app/watch to your docker run command
3. In Scriberr settings, set the watch folder to /app/watch
4. Drop an audio file into ~/scriberr/watch — it gets transcribed automatically

I pair this with Syncthing on my phone: I record a voice memo, Syncthing syncs it to the watch folder, and Scriberr transcribes it. Completely hands-free.

Troubleshooting

What's Next?

Scriberr has a REST API too — you can integrate it with n8n, Home Assistant, or any system that can make HTTP requests. The API docs are at scriberr.app/api.

The project is currently in a maintenance pause as the creator was affected by layoffs, but the software works perfectly as-is. If you find it useful, consider supporting them on Ko-fi or contributing to the codebase.