Automatic Speech Recognition
GGUF
VibeVoice
ggml
ASR
quantization
cpu-inference
bitnet
conversational
Instructions to use P2Enjoy/VibeVoice-ASR-BitNet-slim with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use P2Enjoy/VibeVoice-ASR-BitNet-slim with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf P2Enjoy/VibeVoice-ASR-BitNet-slim # Run inference directly in the terminal: llama cli -hf P2Enjoy/VibeVoice-ASR-BitNet-slim
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf P2Enjoy/VibeVoice-ASR-BitNet-slim # Run inference directly in the terminal: llama cli -hf P2Enjoy/VibeVoice-ASR-BitNet-slim
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf P2Enjoy/VibeVoice-ASR-BitNet-slim # Run inference directly in the terminal: ./build/bin/llama cli -hf P2Enjoy/VibeVoice-ASR-BitNet-slim
Use Docker
docker model run hf.co/P2Enjoy/VibeVoice-ASR-BitNet-slim
- Llama new
- LM Studio
- Jan
- Ollama
How to use P2Enjoy/VibeVoice-ASR-BitNet-slim with Ollama:
ollama run hf.co/P2Enjoy/VibeVoice-ASR-BitNet-slim
- Unsloth Desktop
- Docker Model Runner
How to use P2Enjoy/VibeVoice-ASR-BitNet-slim with Docker Model Runner:
docker model run hf.co/P2Enjoy/VibeVoice-ASR-BitNet-slim
- Lemonade
How to use P2Enjoy/VibeVoice-ASR-BitNet-slim with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull P2Enjoy/VibeVoice-ASR-BitNet-slim
Run and chat with the model
lemonade run user.VibeVoice-ASR-BitNet-slim-{{QUANT_TAG}}List all available models
lemonade list
- Atomic Chat
Accuracy in the open: four-engine WER table incl. whisper; speed as engine ratios
Browse files
README.md
CHANGED
|
@@ -41,16 +41,15 @@ through the Q6_K copy. That is the only numerical change, and it also removes
|
|
| 41 |
|
| 42 |
## Accuracy
|
| 43 |
|
| 44 |
-
|
| 45 |
-
greedy decoding, scored corpus-level with digits and years spelled out on both sides
|
| 46 |
-
(methodology and scripts:
|
| 47 |
-
[bench/README.md](https://github.com/martinobettucci/VibeASR-bitnet.cpp/blob/claude/asr-cpu-optimization-cztnh9/bench/README.md)).
|
| 48 |
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
|
|
|
|
|
|
|
| 54 |
|
| 55 |
| Corpus / language | Released (1.70 GB) | **Slim (1.23 GB)** | Δ |
|
| 56 |
|:--|--:|--:|--:|
|
|
@@ -63,15 +62,35 @@ binary.
|
|
| 63 |
| **FLEURS corpus** | **13.95** | **14.25** | **+0.30** |
|
| 64 |
| MLS French (read speech) | 21.97 | 22.31 | +0.34 |
|
| 65 |
|
| 66 |
-
Four of six
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
The runtime also ships opt-in decoder-level hotword biasing (`--hotwords`,
|
| 77 |
token-trie logit boosting): on FLEURS-French with oracle terms it recovers ~4 WER
|
|
@@ -88,11 +107,14 @@ languages degrade sharply; this repack does not change language coverage.
|
|
| 88 |
|
| 89 |
## Speed
|
| 90 |
|
| 91 |
-
Documented in the runtime repo, not here — it is a property of the code, and
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
|
|
|
|
|
|
|
|
|
| 96 |
|
| 97 |
➡️ **[martinobettucci/VibeASR-bitnet.cpp](https://github.com/martinobettucci/VibeASR-bitnet.cpp/tree/claude/asr-cpu-optimization-cztnh9)** — "CPU optimisation on AVX-512"
|
| 98 |
|
|
|
|
| 41 |
|
| 42 |
## Accuracy
|
| 43 |
|
| 44 |
+
Two measurements, two questions.
|
|
|
|
|
|
|
|
|
|
| 45 |
|
| 46 |
+
**1. Does the slim repack cost accuracy?** FLEURS test clips (24 per language)
|
| 47 |
+
plus a Multilingual LibriSpeech French anchor, greedy decoding, scored
|
| 48 |
+
corpus-level with digits and years spelled out on both sides (methodology and
|
| 49 |
+
scripts:
|
| 50 |
+
[bench/README.md](https://github.com/martinobettucci/VibeASR-bitnet.cpp/blob/claude/asr-cpu-optimization-cztnh9/bench/README.md)).
|
| 51 |
+
Both models on one runtime build (`094b53c`, portable AVX2+VNNI, deterministic —
|
| 52 |
+
identical transcripts at any thread count):
|
| 53 |
|
| 54 |
| Corpus / language | Released (1.70 GB) | **Slim (1.23 GB)** | Δ |
|
| 55 |
|:--|--:|--:|--:|
|
|
|
|
| 62 |
| **FLEURS corpus** | **13.95** | **14.25** | **+0.30** |
|
| 63 |
| MLS French (read speech) | 21.97 | 22.31 | +0.34 |
|
| 64 |
|
| 65 |
+
Four of six languages produce **identical transcripts**; the corpus delta comes
|
| 66 |
+
almost entirely from French, the model's weakest language, where ±1–2 points is
|
| 67 |
+
sampling noise on ~700 reference words. Practical read: **no measurable accuracy
|
| 68 |
+
cost** for the 27% size reduction. The MLS row is the register check: French drops
|
| 69 |
+
from ~35% (encyclopedic FLEURS) to ~22% (audiobooks) for both models — corpus
|
| 70 |
+
difficulty, not quantisation damage.
|
| 71 |
+
|
| 72 |
+
**2. Where does the model sit in the landscape?** A second, independent suite —
|
| 73 |
+
100 clips drawn across the six languages plus the MLS-French anchor, all engines
|
| 74 |
+
back-to-back on one host at 4 threads, later runtime build, whisper given each
|
| 75 |
+
clip's language code (generous: this model runs unhinted):
|
| 76 |
+
|
| 77 |
+
| Engine / weights | de | en | es | fr | fr-MLS | it | pt | **all** |
|
| 78 |
+
|:--|--:|--:|--:|--:|--:|--:|--:|--:|
|
| 79 |
+
| Upstream engine, original 1.70 GB | 16.7 | 6.6 | 5.8 | 32.7 | 21.6 | 8.3 | 10.8 | **15.75** |
|
| 80 |
+
| **This repo (slim), fork engine** | 13.9 | 7.2 | 5.8 | 35.4 | 21.2 | 10.2 | 10.5 | **16.03** |
|
| 81 |
+
| whisper.cpp small q5_1 | 8.5 | 4.6 | 7.1 | 12.9 | 16.7 | 6.6 | 8.0 | **9.92** |
|
| 82 |
+
| whisper.cpp large-v3-turbo q5_0 | 4.1 | 3.7 | 4.0 | 3.3 | 10.3 | 3.3 | 4.3 | **5.12** |
|
| 83 |
+
|
| 84 |
+
Read it straight: slim-vs-original is again parity (+0.28, per-language scatter in
|
| 85 |
+
both directions — German improves under the fork's overflow-fixed kernels, Italian
|
| 86 |
+
regresses), while whisper models beat this architecture on accuracy at this corpus
|
| 87 |
+
register. What this stack offers instead is speed shape and features: compute
|
| 88 |
+
proportional to clip length where whisper always encodes a fixed 30 s window (this
|
| 89 |
+
engine is 1.76× *faster* than whisper-small below 8 s of audio and 3.9× faster
|
| 90 |
+
than turbo overall), native segment/speaker JSON, and decoder-level hotword
|
| 91 |
+
biasing. Speed ratios and their methodology live in the
|
| 92 |
+
[runtime repo](https://github.com/martinobettucci/VibeASR-bitnet.cpp/tree/claude/asr-cpu-optimization-cztnh9)
|
| 93 |
+
and are not duplicated here.
|
| 94 |
|
| 95 |
The runtime also ships opt-in decoder-level hotword biasing (`--hotwords`,
|
| 96 |
token-trie logit boosting): on FLEURS-French with oracle terms it recovers ~4 WER
|
|
|
|
| 107 |
|
| 108 |
## Speed
|
| 109 |
|
| 110 |
+
Documented in the runtime repo, not here — it is a property of the code, and
|
| 111 |
+
absolute RTF is a property of the host, so speed is published as **ratios**
|
| 112 |
+
between engines measured back-to-back: **2.56× faster than the upstream runtime**
|
| 113 |
+
(paired per-clip median, 100 clips, identical weights class), parity with
|
| 114 |
+
whisper-small on long clips and 1.76× faster below 8 s. Achieved via VNNI kernels
|
| 115 |
+
with a packed-B INT8 GEMM (fused dequantisation epilogue), conv-as-GEMM
|
| 116 |
+
downsampling, and a layout-native depthwise convolution. Tables, profiler
|
| 117 |
+
methodology, and full reproduction scripts:
|
| 118 |
|
| 119 |
➡️ **[martinobettucci/VibeASR-bitnet.cpp](https://github.com/martinobettucci/VibeASR-bitnet.cpp/tree/claude/asr-cpu-optimization-cztnh9)** — "CPU optimisation on AVX-512"
|
| 120 |
|