AI/ML●●●Banger
Built speech models for Europe, accidentally topped Open-ASR in English
Processes one hour of audio in under three seconds while topping Open ASR Leaderboard.
WizardryDark Horse
raoulritter
202mo ago
Back to browse
150M Mandarin transcription model with real-time metadata detection
by ksingla025·Jun 18, 2026·1 point·0 comments
AI Analysis
●●SolidNiche GemBig Brain
Dual-head Citrinet fine-tune beats running Whisper plus a separate classifier.
Strengths
- •Single forward pass outputs transcription and speaker metadata simultaneously
- •ONNX exports included for straightforward deployment across platforms
- •94.2% tag accuracy on age, gender, and dialect classification is solid
Weaknesses
- •Only 60 hours of training data limits generalization to diverse speakers
- •Mandarin-only scope restricts broader applicability beyond Chinese markets
Category
Target Audience
Developers building Mandarin speech applications needing speaker metadata
Similar To
Whisper · NVIDIA NeMo · Azure Speech API
Similar Projects
AI/ML●Mid
I Added Support for Qwen3-ASR and Qwen3 ForcedAligner in WhisperX
Adds Qwen3 model support to WhisperX for users wanting alternatives.
Niche Gem
mahfouz22
212mo ago
AI/ML●●●Banger
Transcribe-Critic – Merge transcript sources for stronger transcript
Textual-criticism approach to transcript merging beats single-model Whisper on accuracy alone.
Big BrainZero to One
ringger
213mo ago
AI/ML●●●Banger
Overlapping Speaker Transcription Model
Transcribes overlapping speakers in a single pass without needing separate diarization steps.
WizardryBig BrainSolve My Problem
mcgov
101mo ago
Infrastructure●●Solid
ZSE – Single-file LLM engine with dual INT4 kernels
INT4 inference engine beats llama.cpp on VRAM, but competing against established tools.
WizardryShip It
zyoralabs
103mo ago
AI/ML●Mid
Transcribe, analyse, and chat with your voice conversations
Open-source Otter.ai clone with local LLMs, but meeting transcription is crowded.
Ship ItSolve My Problem
nezhar
102mo ago