An offline speech transcript normalizer and local AI SEO content generator powered by superwhisper/s1-mini GGUF and Ollama to clean raw speech dictation and generate structured SEO content packages locally.
| Step 1: Input | Step 2: s1-mini Clean | Step 3: Local AI SEO |
|---|---|---|
| Raw unpunctuated voice ASR transcript | s1-mini 0.6B normalizes text locally in Ollama | Stage 2 Local AI generates SEO title, tags, and social posts |
- 🦙 Native Ollama Support: Run s1-mini GGUF directly inside Ollama without requiring external servers.
- 🔒 100% Offline Privacy: Clean voice transcripts and build SEO content locally with zero cloud API fees.
- 🔄 2-Stage AI Pipeline: Stage 1 cleans raw transcripts with
s1-mini; Stage 2 builds complete SEO packages using any local LLM. - 🤖 Dynamic Stage 2 Model Selector: Select or change installed Ollama models (
llama3.2,qwen2.5,mistral, etc.) directly in the Streamlit UI. - 🎛️ 3-Axis Steering Control: Precise control over register styling (casual to formal), structure (prose or lists), and context (general or email).
- 🖥️ Streamlit Frontend: Interactive web interface to clean speech dictations and generate instant SEO summaries.
The .gguf weight files are excluded from this git repository due to file size. Download your preferred quantization from Hugging Face and place it in the models/ directory before building the model:
| Quantization | File Name | Size | Download Link |
|---|---|---|---|
| Q4_K_M (Recommended) | s1-mini-q4_k_m.gguf |
~462 MB | 📥 Download Q4_K_M GGUF |
| FP16 (Full Precision) | s1-mini-f16.gguf |
~1.4 GB | 📥 Download FP16 GGUF |
🔗 Official Hugging Face Repository: superwhisper/s1-mini-GGUF
- Stage 1 Model:
superwhisper/s1-mini(0.6B Q4_K_M GGUF) - Stage 2 Model: Any Local Ollama LLM (
llama3.2:3b,qwen2.5,mistral, etc.) - Local Engine: Ollama Native Server (
http://localhost:11434) - Frontend App: Python & Streamlit
Download s1-mini-q4_k_m.gguf from the link above and place it in the models/ folder:
s1-mini/models/s1-mini-q4_k_m.gguf
Create and register the s1-mini model in Ollama using the included Modelfile:
ollama create s1-mini -f ./models/Modelfileollama listpip install -r requirements.txt
streamlit run main.pys1-mini/
├── main.py
├── requirements.txt
├── README.md
├── .gitignore
└── models/
└── Modelfile
- Voice-Driven Content Creation: Dictate raw thoughts on a microphone and automatically output clean articles and SEO metadata.
- Offline Technical Bug Logging: Convert developer voice dictations into clean issue reports and social update posts.
- Private Meeting Summaries: Process sensitive business discussions locally with zero third-party cloud data transfer.
- Podcast Episode SEO Packages: Transform raw episode voice notes into SEO titles, meta descriptions, and X/LinkedIn posts.
- Voice-to-Email Drafting: Clean informal speech transcripts and structure them into formal email drafts.
- 🎙️ One-Click Audio Recording: Integrate local Whisper microphone recording directly inside the Streamlit frontend.
- 🌐 Multi-Language Normalization: Extend normalization steering to Spanish, German, and French speech transcripts.
- 📊 Automatic Edit Distance Score: Display real-time token edit distance comparing raw vs. s1-mini cleaned text.
- 🔌 CMS & Social Auto-Poster: Export generated SEO packages directly to WordPress, Ghost, and LinkedIn drafts.
- 🏷️ Dynamic Keyword Density Visualizer: Graph keyword frequency and search optimization density for output posts.
s1-mini superwhisper qwen2.5-0.6b speech-recognition transcript-normalizer ollama gguf local-ai streamlit youtube-seo dictation-cleaner privacy-first ai-automation