Audio-to-Audio
Transformers
Korean
audio
speech
audio-language-model
speech-understanding
spoken-dialogue
speech-generation
voice-activity-detection
korean
alm
Instructions to use skt/A.X-K2-ALM with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use skt/A.X-K2-ALM with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("skt/A.X-K2-ALM", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Fix demo video playback across browsers/devices (faststart remux, MP4 brand fix, stable CDN embed with fallback)
Browse files- README.md +10 -2
- assets/health_demo.mp4 +2 -2
- assets/vad_demo.mp4 +1 -1
README.md
CHANGED
|
@@ -119,12 +119,20 @@ Averaged over the ten KVoiceBench subsets, A.X K2 ALM scores **64.27 — 95% of
|
|
| 119 |
**1. Natural Voice Conversation**
|
| 120 |
A fluid, natural spoken dialogue with the model.
|
| 121 |
|
| 122 |
-
<video controls
|
|
|
|
|
|
|
|
|
|
|
|
|
| 123 |
|
| 124 |
**2. Context-Aware Turn-Taking**
|
| 125 |
The context-aware VAD ignores listener back-channels (e.g., "음", "아") and waits through mid-sentence hesitations — so it neither falsely stops nor cuts in, responding only once the user has actually finished.
|
| 126 |
|
| 127 |
-
<video controls
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
|
| 129 |
## Availability
|
| 130 |
|
|
|
|
| 119 |
**1. Natural Voice Conversation**
|
| 120 |
A fluid, natural spoken dialogue with the model.
|
| 121 |
|
| 122 |
+
<video controls playsinline preload="metadata" width="720">
|
| 123 |
+
<source src="https://cdn-uploads.huggingface.co/production/uploads/noauth/5CaJJY7leg7W8GWfDwyJs.mp4" type="video/mp4">
|
| 124 |
+
<source src="https://huggingface.co/skt/A.X-K2-ALM/resolve/main/assets/health_demo.mp4" type="video/mp4">
|
| 125 |
+
<a href="https://huggingface.co/skt/A.X-K2-ALM/resolve/main/assets/health_demo.mp4?download=true">Download the demo video</a>
|
| 126 |
+
</video>
|
| 127 |
|
| 128 |
**2. Context-Aware Turn-Taking**
|
| 129 |
The context-aware VAD ignores listener back-channels (e.g., "음", "아") and waits through mid-sentence hesitations — so it neither falsely stops nor cuts in, responding only once the user has actually finished.
|
| 130 |
|
| 131 |
+
<video controls playsinline preload="metadata" width="720">
|
| 132 |
+
<source src="https://cdn-uploads.huggingface.co/production/uploads/noauth/htPKhczYGj4Nnh4jYKddh.mp4" type="video/mp4">
|
| 133 |
+
<source src="https://huggingface.co/skt/A.X-K2-ALM/resolve/main/assets/vad_demo.mp4" type="video/mp4">
|
| 134 |
+
<a href="https://huggingface.co/skt/A.X-K2-ALM/resolve/main/assets/vad_demo.mp4?download=true">Download the demo video</a>
|
| 135 |
+
</video>
|
| 136 |
|
| 137 |
## Availability
|
| 138 |
|
assets/health_demo.mp4
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ac6c77aab1cca4ec6e4bec5d18f132b730790dd1e006ddb2baf1dc9b09bdc5b
|
| 3 |
+
size 19403082
|
assets/vad_demo.mp4
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 13294736
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:48a077662c2e6b1d5350f9199cdcd00618dc0606570316adea15f1c8395afc05
|
| 3 |
size 13294736
|