AI Music Detection¶

Detect AI-generated music.

Endpoint¶

Method	URI	Description
POST	`/api/v1/ai-detection/{model_name}`	Detect AI-generated music

Path Parameters¶

Name	Type	Required	Description
`model_name`	string	Yes	Model to use: `lite`, `standard`, `pro`, `pro-v2`

Request Body¶

Field	Type	Required	Description
`file`	binary	Yes	Audio file to analyze (mp3, wav, flac, m4a, aac, ogg)

Models¶

Model	Processing Time	Description
`lite`	30-40 seconds	Single model, fast detection
`standard`	30-40 seconds	Model ensemble, balanced accuracy
`pro`	~1 minute	Model ensemble with detailed classification
`pro-v2`	~15 seconds	AI tracking with role-based detection

Request Example¶

cURL¶

curl https://platform.mippia.com/api/v1/ai-detection/standard \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -X POST \
  -F "file=@/path/to/audio.mp3"

Python¶

import requests

url = "https://platform.mippia.com/api/v1/ai-detection/standard"
headers = {
    "Authorization": "Bearer YOUR_API_KEY"
}
files = {
    "file": open("/path/to/audio.mp3", "rb")
}

response = requests.post(url, headers=headers, files=files)
print(response.json())

Response (Initial)¶

{
  "task_id": "task_20251204052920_J8uNdq5z",
  "status": "pending",
  "filepath": "uploads/task_20251204052920_J8uNdq5z.mp3",
  "model": "standard",
  "created_at": "2025-12-04T05:29:20Z"
}

Callback Response (Processing)¶

{
  "task_id": "task_20251204052920_J8uNdq5z",
  "status": "processing",
  "result": null,
  "model_type": "standard"
}

Callback Response (Completed - Standard)¶

{
  "task_id": "task_20251204052920_J8uNdq5z",
  "status": "success",
  "model_type": "standard",
  "completed_at": "2025-12-04T05:30:15Z",
  "result": {
    "audio_filename.mp3": {
      "final_result": {
        "prediction": "real",
        "confidence": 0.9382
      },
      "model_results": [
        {
          "overall_analysis": {
            "prediction": "real",
            "confidence": 0.923
          },
          "config": {
            "model_id": "model_0",
            "analysis_focus": "Waveform & Melody Pattern",
            "task": "Real/Fake Binary Classification",
            "num_classes": 2,
            "labels": ["real", "fake"]
          }
        },
        {
          "segment_analysis": {
            "prediction": ["real", "real", "real", "real"],
            "confidence": [0.891, 0.912, 0.887, 0.903]
          },
          "overall_analysis": {
            "prediction": "real",
            "confidence": 0.945
          },
          "config": {
            "model_id": "model_1",
            "analysis_focus": "Waveform & Melody Pattern",
            "task": "Real/Fake Binary Classification",
            "num_classes": 2,
            "labels": ["real", "fake"]
          }
        },
        {
          "segment_analysis": {
            "prediction": ["real", "real", "real", "real"],
            "confidence": [0.876, 0.901, 0.889, 0.912]
          },
          "overall_analysis": {
            "prediction": "real",
            "confidence": 0.934
          },
          "config": {
            "model_id": "model_2",
            "analysis_focus": "Waveform & Melody Pattern",
            "task": "Real/Fake Binary Classification",
            "num_classes": 2,
            "labels": ["real", "fake"]
          }
        },
        {
          "segment_analysis": {
            "prediction": ["real", "real", "real", "real"],
            "confidence": [0.902, 0.918, 0.895, 0.921]
          },
          "overall_analysis": {
            "prediction": "real",
            "confidence": 0.951
          },
          "config": {
            "model_id": "model_3",
            "analysis_focus": "Waveform & Melody Pattern",
            "task": "Real/Fake Binary Classification",
            "num_classes": 2,
            "labels": ["real", "fake"]
          }
        }
      ]
    }
  }
}

Callback Response (Completed - Pro)¶

Pro model includes additional classifiers for detailed analysis:

{
  "task_id": "task_20251204052920_J8uNdq5z",
  "status": "success",
  "model_type": "pro",
  "completed_at": "2025-12-04T05:30:45Z",
  "result": {
    "audio_filename.mp3": {
      "final_result": {
        "prediction": "fake",
        "confidence": 0.8654
      },
      "model_results": [
        {
          "overall_analysis": {
            "prediction": "fake",
            "confidence": 0.876
          },
          "config": {
            "model_id": "model_0",
            "analysis_focus": "Waveform & Melody Pattern",
            "task": "Real/Fake Binary Classification",
            "num_classes": 2,
            "labels": ["real", "fake"]
          }
        },
        {
          "segment_analysis": {
            "prediction": ["fake", "fake", "ai_cover", "fake"],
            "confidence": [0.712, 0.834, 0.623, 0.789]
          },
          "overall_analysis": {
            "prediction": "fake",
            "confidence": 0.823
          },
          "config": {
            "model_id": "model_4",
            "analysis_focus": "Mixing & Audio Effects",
            "task": "Real/AI-Cover/Fake 3-Class",
            "num_classes": 3,
            "labels": ["real", "ai_cover", "fake"]
          }
        },
        {
          "segment_analysis": {
            "prediction": ["suno_v4", "suno_v4", "suno_v4_5", "suno_v4"],
            "confidence": [0.534, 0.612, 0.489, 0.567]
          },
          "overall_analysis": {
            "prediction": "suno_v4",
            "confidence": 0.634
          },
          "config": {
            "model_id": "model_5",
            "analysis_focus": "Waveform & Melody Pattern",
            "task": "Detailed Fake Source Classification",
            "num_classes": 6,
            "labels": ["real", "suno_v4", "suno_v4_5", "suno_v4_5_plus", "suno_v5", "other"]
          }
        },
        {
          "segment_analysis": {
            "prediction": ["fake", "fake", "fake", "fake"],
            "confidence": [0.891, 0.912, 0.878, 0.901]
          },
          "overall_analysis": {
            "prediction": "fake",
            "confidence": 0.912
          },
          "config": {
            "model_id": "model_6",
            "analysis_focus": "Waveform & Melody Pattern",
            "task": "Real/AI-Cover/Fake 3-Class",
            "num_classes": 3,
            "labels": ["real", "ai_cover", "fake"]
          }
        }
      ]
    }
  }
}

Callback Response (Completed - Pro-V2)¶

Pro-V2 model uses role-based AI detection (composer, vocalist, engineer) and AI model identification. Unlike other models, the result object contains final_result and model_results directly (no filename key wrapper).

{
  "task_id": "task_20251204052920_J8uNdq5z",
  "status": "success",
  "model_type": "pro-v2",
  "completed_at": "2025-12-04T05:30:20Z",
  "result": {
    "final_result": {
      "prediction": "fake",
      "confidence": 0.99
    },
    "model_results": [
      {
        "segment_analysis": {
          "prediction": ["fake", "fake", "fake", "fake", "fake", "fake"],
          "confidence": [0.9735, 0.9764, 0.9788, 0.97, 0.9848, 0.9775]
        },
        "overall_analysis": {
          "prediction": "fake",
          "confidence": 0.9768
        },
        "config": {
          "model_id": "2label_stage1",
          "analysis_focus": "Waveform & Melody Pattern",
          "task": "Real/Fake Binary Classification",
          "num_classes": 2,
          "labels": ["real", "fake"]
        }
      },
      {
        "overall_analysis": {
          "prediction": "fake",
          "confidence": 0.99
        },
        "config": {
          "model_id": "2label_stage2",
          "analysis_focus": "Waveform & Melody Pattern",
          "task": "Real/Fake Binary Classification",
          "num_classes": 2,
          "labels": ["real", "fake"]
        }
      },
      {
        "segment_analysis": {
          "prediction": [
            ["composer", "vocalist", "engineer"],
            ["composer", "vocalist", "engineer"],
            ["composer", "vocalist", "engineer"],
            ["real"],
            ["composer", "vocalist", "engineer"],
            ["real"]
          ],
          "confidence": [
            {"composer": 0.99, "vocalist": 0.99, "engineer": 0.99},
            {"composer": 0.99, "vocalist": 0.99, "engineer": 0.99},
            {"composer": 0.99, "vocalist": 0.99, "engineer": 0.99},
            {"composer": 0.01, "vocalist": 0.03, "engineer": 0.03},
            {"composer": 0.99, "vocalist": 0.99, "engineer": 0.99},
            {"composer": 0.01, "vocalist": 0.01, "engineer": 0.01}
          ]
        },
        "overall_analysis": {
          "prediction": ["composer", "vocalist", "engineer"],
          "confidence": {
            "composer": 0.99,
            "vocalist": 0.99,
            "engineer": 0.99
          }
        },
        "config": {
          "model_id": "agent_classifier",
          "analysis_focus": "Role-based AI Detection",
          "task": "Per-agent AI Classification",
          "labels": ["composer", "vocalist", "engineer"]
        }
      },
      {
        "overall_analysis": {
          "prediction": "suno",
          "confidence": 0.99,
          "all_probs": {
            "acestep": 0.01,
            "lyria-pro3": 0.01,
            "mureka": 0.01,
            "suno": 0.99,
            "udio": 0.01
          }
        },
        "config": {
          "model_id": "fake_model_classifier",
          "analysis_focus": "AI Model Identification",
          "task": "Fake Model Classification",
          "num_classes": 5,
          "labels": ["acestep", "lyria-pro3", "mureka", "suno", "udio"]
        }
      }
    ]
  }
}

Result Fields¶

Field	Type	Description
`task_id`	string	Unique task identifier
`status`	string	Task status: `pending`, `processing`, `success`, `failure`
`model_type`	string	Model used for detection
`completed_at`	string	ISO 8601 completion timestamp
`result`	object	Detection results keyed by filename

Result Structure¶

For lite, standard, pro: the result is an object where the key is the audio filename and the value contains final_result and model_results.

For pro-v2: the result object contains final_result and model_results directly (no filename key).

Field	Type	Description
`final_result`	object	Aggregated prediction from all Real/Fake binary classification models
`model_results`	array	Individual results from each model in the ensemble

Final Result Fields¶

Field	Type	Description
`prediction`	string	Final aggregated prediction: `real` or `fake`
`confidence`	float	Final confidence score (0.0 - 1.0)

Model Result Fields¶

Field	Type	Description
`segment_analysis`	object	Results from analyzing individual segments (optional)
`overall_analysis`	object	Final prediction for the entire audio
`config`	object	Model configuration and metadata

Config Fields¶

Field	Type	Description
`model_id`	string	Unique model identifier (e.g., `model_0`, `model_1`)
`analysis_focus`	string	What the model analyzes (see Analysis Focus below)
`task`	string	Classification task type
`num_classes`	integer	Number of output classes
`labels`	array	Possible prediction labels

Analysis Focus¶

Focus	Description
Waveform & Melody Pattern	Analyzes audio waveform characteristics and melodic patterns
Mixing & Audio Effects	Analyzes mixing techniques and audio effect signatures
Role-based AI Detection	Analyzes AI involvement by musical role (composer, vocalist, engineer)
AI Model Identification	Identifies which AI model was used to generate the audio

Analysis Result Fields¶

Field	Type	Description
`segment_analysis`	object	Results from analyzing individual segments of the audio
`segment_analysis.prediction`	array	Prediction for each audio segment
`segment_analysis.confidence`	array	Confidence score for each segment
`overall_analysis`	object	Final prediction considering the entire song structure
`overall_analysis.prediction`	string	Final aggregated prediction
`overall_analysis.confidence`	float	Final confidence score (0.0 - 1.0)

Classification Labels¶

Task	Labels
Real/Fake Binary	`real`, `fake`
3-Class	`real`, `ai_cover`, `fake`
Detailed Fake Source	`real`, `suno_v4`, `suno_v4_5`, `suno_v4_5_plus`, `suno_v5`, `other`
Role-based AI	`real`, `composer`, `vocalist`, `engineer`
AI Model Identification	`real`, `acestep`, `lyria-pro3`, `mureka`, `suno`, `udio`

Notes¶

Supported formats: mp3, wav, flac, m4a, aac, ogg
Final Result: Weighted average of all Real/Fake binary classification models for a single consolidated prediction
Segment Analysis: Analyzes each segment of the audio independently to detect localized AI artifacts
Overall Analysis: Considers the entire song structure and aggregates segment results for final prediction
Model Results: Each element in the model_results array represents one model’s analysis
Pro-V2 Role Detection: The agent classifier detects AI involvement by musical role (composer, vocalist, engineer). For segment-level analysis, prediction is an array of arrays where each segment contains detected roles (or ["real"] if no AI detected). Confidence provides per-role probabilities for each segment.
Pro-V2 AI Model Identification: The fake model classifier identifies which AI model likely generated the audio (acestep, lyria-pro3, mureka, suno, udio). Probabilities are scaled by the overall fake confidence — if the track is predicted as real, all model probabilities will be near zero.