LingoBird logoLingoBird
Hybrid Offline + Cloud Architecture

Neural Speech & AI On Your Terms

LingoBird runs neural voices, microphone transcription, and translation models directly inside Windows via WebAssembly & ONNX — and optionally pairs with free Gemini/Gemma API or Local Ollama.

AI Tutor & Speech Pipeline in Action

Drop your AI Tutor screen recordings into public/videos/ to feature them automatically here.

Official Walkthrough

How to Use the AI Teacher in Lingobird ?

YouTube

Watch a complete live session: synchronized Piper neural voices stream tutor speech in real-time, Whisper local speech recognition transcribes pronunciation with zero cloud lag, and Google Gemma 4 26B or local Ollama powers free, private conversational practice.

Synchronized Voice Streaming

Piper neural voices read replies with live word-by-word streaming, phonetic subtitles, and 0.8×/1.0× playback speed toggles.

Demonstrated in Video

Whisper Speech-to-Text

Transcribe your spoken pronunciation on-device with zero cloud lag, or tap any unfamiliar tutor word for instant translation and flashcard capture.

Demonstrated in Video

Gemma 4 26B & Local Ollama

Connect Google AI Studio (14,400 free daily queries) or run Ollama models 100% offline for private, unlimited voice tutoring.

Demonstrated in Video
Subscribe to the official LingoBird channel for new feature walk-throughs & language tips.Visit @lingobird_app
AI Tutor & Grading Models

Free Google AI Studio or 100% Local Ollama

View setup walkthrough
Google AI StudioRecommended · 14.4K free req/day

Google Gemma 4 26B / 31B

gemma-4-26b-a4b-it · gemma-4-31b-it

Generous daily quota with rich multilingual reasoning & roleplay quality.

Google AI StudioLow-Latency Cloud

Google Gemini 2.5 Flash & Pro

gemini-2.5-flash · gemini-2.5-pro

Fast daily driver for instant answer grading, story generation & deep grammar breakdowns.

Local Ollama100% Offline Recommended

Ollama (gemma3:4b)

gemma3:4b (localhost:11434)

Runs completely locally on Windows for zero-cost, zero-cloud-latency tutoring.

Local OllamaAuto-Detected

Llama 3.2 · Mistral · DeepSeek R1 · Qwen 2.5

llama3.2 · mistral · deepseek-r1:8b · qwen2.5:3b

One-click local Ollama scanner adds any model already pulled on your machine.

Offline Voices, Whisper STT & AI Models SettingsLingoBird · Windows
Offline Voices, Whisper STT & AI Models Settings
Piper neural voice manager, Whisper local STT models, Helsinki-NLP translation packs, Gemini/Gemma API setup, and Ollama scanner.

20 Offline Piper Neural Voices

Female & Male ONNX voices across 10 languages (+ Japanese OpenJTalk) cached locally

🇺🇸 English (US)
Lessac (F) · Ryan (M)
~60 MB
🇪🇸 Spanish
Sharvard (F) · DaveFX (M)
~60–73 MB
🇫🇷 French
Siwis (F) · Tom (M)
~60 MB
🇩🇪 German
Ramona (F) · Thorsten (M)
~60 MB
🇮🇹 Italian
Paola (F) · Riccardo (M)
~27–60 MB
🇵🇹 Portuguese
Tugão (F) · Faber (M)
~60 MB
🇷🇺 Russian
Irina (F) · Denis (M)
~60 MB
🇨🇳 Mandarin Chinese
Huayan (F) · Chaowen (M)
~60 MB
🇮🇳 Hindi
Priyamvada (F) · Pratham (M)
~60 MB
🇯🇵 Japanese
Hi-Fi Captain F & M (OpenJTalk)
~73 MB

Speech-to-Text (STT)

Whisper Tiny Multilingual (On-Device)~39 MB · <150 MB RAM

Fastest int8 ONNX WebAssembly transcription for 99+ languages right inside Windows.

Whisper Base Multilingual (On-Device)~73 MB · <220 MB RAM

97% accent & grammar precision with stronger tolerance for natural learner speech.

Whisper Tiny English & Groq Cloud STT~39 MB local or Cloud fallback

Dedicated English pronunciation precision offline or ultra-fast Groq Whisper API option.

Offline OPUS-MT Translators

12 quantized neural pairs (~31–39 MB each) run offline via Transformers.js inside the Overview Translator and study views.

🇪🇸 ⇄ 🇬🇧 Spanish ⇄ English (2 packs)🇫🇷 ⇄ 🇬🇧 French ⇄ English (2 packs)🇩🇪 ⇄ 🇬🇧 German ⇄ English (2 packs)🇮🇹 → 🇬🇧 Italian → English🇵🇹 → 🇬🇧 Portuguese → English🇷🇺 → 🇬🇧 Russian → English🇨🇳 → 🇬🇧 Chinese → English🇯🇵 → 🇬🇧 Japanese → English🇮🇳 → 🇬🇧 Hindi → English
Audio captured during mic exercises stays in volatile RAM.