File size: 4,440 Bytes
64ed457
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6724d41
64ed457
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
---
language:
  - en
license: apache-2.0
library_name: dotcheck
pipeline_tag: audio-classification
tags:
  - audio-classification
  - ai-detection
  - synthetic-media
  - dasheng
  - apache-2.0
  - dotcheck
  - helmholtz
base_model: mispeech/dasheng-base
base_model_relation: adapter
model-index:
  - name: Helmholtz (inhouse-audio@2)
    results:
      - task:
          type: audio-classification
          name: binary AI-likeness (audio)
        dataset:
          name: DotCheck audio holdout (CodecFake / DFADD / AudioGen)
          type: other
          split: holdout
        metrics:
          - name: mean_P_AI_real
            type: mean_score_real
            value: 0.004
          - name: mean_P_AI_ai
            type: mean_score_ai
            value: 0.984
          - name: balanced_accuracy
            type: balanced_accuracy
            value: 0.986
        source:
          name: audio_gates_audio_v2 / Data.json
          url: https://dotcheck.ai/docs
---

# DotCheck/helmholtz-audio-v2

Apache-2.0 audio AI-likeness head for DotCheck. This repo includes the live .npz head, model card, license, and notices.

| Field | Value |
|-------|--------|
| Hub id | `DotCheck/helmholtz-audio-v2` |
| Wire id | `inhouse-audio@2` |
| Label | Helmholtz |
| Artifact | `dasheng_base_linear_head_audio_v2.npz` |
| Backbone | [`mispeech/dasheng-base`](https://huggingface.co/mispeech/dasheng-base) (Apache-2.0) |
| Protocol | 4 s / center crop / up to 2 windows / max aggregate (serve may single-pass prewindowed clips) |
| Output | `p ∈ [0,1]` |
| Serve | CPU FastAPI `/v1/analyze-audio`; Express Dashboard + video soundtrack fuse |

## Model description

Frozen Dasheng-Base masked audio encoder + logistic head. Scores how AI-like speech, music, and general audio sound. Product **Covenant** seats soundtrack windows with Muybridge frame bags (AU6 fuse; default `max`) when a video has usable audio. Covenant seating is DotCheck-private; this repo is the open audio person only.

**Files in this repo:** `README.md`, `LICENSE`, `NOTICE`, `CITATION.cff`, and the `.npz` head file(s) listed above.

## Architecture

```text
audio bytes
  → mono 16 kHz
  → protocol window(s) (4 s)
  → frozen Dasheng-Base emb → logistic head → p
  → multi-window: max
```

Missing audio head at serve → `/v1/analyze-audio` returns **503** (fail closed).

## Inference

Open weights: the live `.npz` head(s) in this repo (Apache-2.0), for use with the frozen upstream backbone named above. This is not a transformers `AutoModel.from_pretrained("DotCheck/…")` package.

Product scoring: Check/Dashboard or Pro API (below). Leviathan (shared memory and related product path) is not in these files.

```bash
curl -sS -X POST "https://dotcheck-server-c221c1f32c68.herokuapp.com/analyze-audio" \
  -H "Authorization: Bearer dc_YOUR_KEY" \
  -F "audio=@clip.wav"
```

UI: https://dotcheck.ai/check · Dashboard audio: https://dotcheck.ai/dashboard · API: https://dotcheck.ai/api · docs: https://dotcheck.ai/docs

## Training data

| Split | Content |
|-------|---------|
| Fit reals | LibriTTS + FSD50K (+ music-from-FSD) |
| Fit AI | parler-mini + dasheng-audiogen (+ contingency codecfake-fit / dfadd-fit) |
| Holdout speech | CodecFake / DFADD (sha-disjoint) |
| Holdout music/SFX | AudioGen odd-seed partition |

Evidence: `audio_gates_audio_v2.json` · `AUDIO_GATES_OK`.

## Evaluation

### Holdout (public claim SSOT)

| Metric | Target | Measured |
|--------|--------|---------:|
| mean P(AI) \| real | ≤ 0.12 | **0.004** |
| mean P(AI) \| AI | ≥ 0.85 | **0.984** |
| bal_acc @ thr | ≥ 0.90 | **0.986** |

## Intended use

- Standalone audio AI-likeness in DotCheck inference.
- Soundtrack windows fused with video frames in product serve.
- Citation of the holdout table above.

### Out of scope

- Product scoring SLA / Leviathan / FUP via Hub download
- Courtroom deepfake forensics / speaker identification
- Legal determinations

## Limitations

- Protocol windows are short; long-form structure is not modeled.
- Codec / compression / unseen generators can shift scores.
- Public floats are standalone audio gates; fused video bags are not a separate claim table yet.

## License

[`LICENSE`](LICENSE) — Apache License 2.0 for DotCheck heads in this repo. Upstream backbones: see [`NOTICE`](NOTICE).

## Citation

[`CITATION.cff`](CITATION.cff) · wire `inhouse-audio@2` / Helmholtz@2 · https://dotcheck.ai/docs