commit d42781bf4e43a63e2c1611be5594464213006246 Author: jingyaogong Date: Wed May 6 14:18:14 2026 +0800 [init] website diff --git a/images/architecture.jpg b/images/architecture.jpg new file mode 100644 index 0000000..7268755 Binary files /dev/null and b/images/architecture.jpg differ diff --git a/images/image2audio_qualitative.jpg b/images/image2audio_qualitative.jpg new file mode 100644 index 0000000..97a032f Binary files /dev/null and b/images/image2audio_qualitative.jpg differ diff --git a/images/logo.png b/images/logo.png new file mode 100644 index 0000000..4ed1d22 Binary files /dev/null and b/images/logo.png differ diff --git a/images/qual_a2a.jpg b/images/qual_a2a.jpg new file mode 100644 index 0000000..4cb962e Binary files /dev/null and b/images/qual_a2a.jpg differ diff --git a/index.html b/index.html new file mode 100644 index 0000000..b87a866 --- /dev/null +++ b/index.html @@ -0,0 +1,739 @@ + + + + + + MiniMind-O - Train a Tiny Omni Model from Scratch + + + + +
+ +
+ +
+ + +
+ Visitor Badge + GitHub Stars + License + Last Commit +
+ + +
+

+ Train a 0.1B Omni Model from scratch.
+ Text. Image. Audio. Streaming Voice.
+ Hear the World. +

+
+ + +
+
+

0.1B

+

Active Parameters

+
+
+

3O

+

Text · Image · Audio

+
+
+

MTP

+

8-codebook Prediction

+
+
+

24kHz

+

Streaming Voice

+
+
+ + +
✨ What You Get
+
+
+

🎧 Omni Interaction

+

Text, image and audio inputs with streaming speech output.

+
+
+

🧠 Thinker–Talker

+

Multimodal understanding and speech generation are decoupled.

+
+
+

⚡ Multi-Token Prediction

+

Predict Mimi codebooks in parallel for lower streaming latency.

+
+
+

🎙️ Voice Cloning Beta

+

In-context speaker control with reference Mimi codes and CAM++.

+
+
+

🧩 Full Pipeline

+

Code, weights and main training data are released together.

+
+
+

� Readable Baseline

+

A compact Omni system designed for study and modification.

+
+
+ + +
📦 Models
+
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
ModelParametersInference MemoryRelease
minimind-3o-moe312M-A115M~1.5 GB2026.05
minimind-3o115M~1.0 GB2026.05
PyTorch weightsDense / MoETrainingReleased
Transformers formatDense / MoEInferenceReleased
+
+ + +
📰 What's New
+
+
+
+ 🔥 2026-05-05 + − +
+
+
    +
  • MiniMind-O is open-sourced for the first time, with minimind-3o (115M) and minimind-3o-moe (312M-A115M) released.
    • +
  • Thinker–Talker dual-path architecture: Talker uses MTP to predict multi-layer Mimi codes, supporting 24 kHz streaming voice generation and barge-in interruption.
    • +
  • Audio codec uses Mimi (8 codebooks, 12.5 Hz, 24 kHz); Talker uses a shared backbone with lightweight adapters at the codebook interface.
    • +
  • Speech and vision features are encoded by frozen SenseVoice-Small and SigLIP2, then injected into the MiniMind hidden space through two-layer MLP projectors.
    • +
  • Both mini and full training datasets are released; the mini setup can run through the full Thinker–Talker pipeline on a single RTX 3090 in about 2 hours.
    • +
  • Built-in 5 voice prompts and 7 unseen voice prompts, with voice cloning and phone-mode WebUI support.
    • +
+
+
+
+ + +
🎮 Inside MiniMind-O
+
+
+ MiniMind-O Architecture + A2A Qualitative Examples + Image to Audio Examples +
+
+ + +
+

🎯 Try It Online

+
+ + 💬 Online Demo + + + 🎬 Video Intro + + + 📄 Technical Report + + + 🔗 GitHub + + + 🤗 HuggingFace + +
+ +
+ +

📦 Get the Code

+
+ + GitHub Repo + + + 🤗 HuggingFace + + + ModelScope + +
+
+ + +
+

💡 Why MiniMind-O?

+
    +
  • Build a complete Omni pipeline around a 0.1B-scale model
    • +
  • Understand Thinker–Talker decoupling from readable code
    • +
  • Study Mimi codebook prediction and streaming voice generation
    • +
  • Experiment with in-context voice cloning in a small system
    • +
  • Reproduce training with released weights and main datasets
    • +
  • Use it as a compact baseline for education and research
    • +
+

+ 💭 "A tiny Omni baseline, small enough to inspect and complete enough to learn from." +

+
+ + +
+

+ MiniMind-O is a compact Omni baseline for study and reproduction.
+ Text, image, audio input · streaming voice output
+ Apache-2.0 License · + GitHub +

+
+
+ + + + + + + \ No newline at end of file