--- name: resemble-detect category: analytics description: Deepfake detection and media safety — detect AI-generated audio, images, video, and text, trace synthesis sources, apply watermarks, verify speaker identity, and analyze media intelligence using Resemble AI license: Apache-2.0 (see LICENSE) --- # Resemble Detect — Deepfake Detection & Media Safety Analyze audio, image, video, and text for synthetic manipulation, AI-generated content, watermarks, speaker identity, and media intelligence using the Resemble AI platform. ## Core Principle — THE IRON LAW **"NEVER DECLARE MEDIA AS REAL OR FAKE WITHOUT A COMPLETED DETECTION RESULT."** Do not guess, infer, or speculate about media authenticity. Every authenticity claim must be backed by a completed Resemble detect job with a returned `label`, `score`, and `status: "completed"`. If the detection is still `processing`, wait. If it `failed`, say so — do not substitute your own judgment. ## When to Use Use this skill whenever the user's request involves any of these: - Checking if audio, video, image, or text is AI-generated or manipulated - Detecting deepfakes in any media format - Verifying media authenticity or provenance - Identifying which AI platform synthesized audio (source tracing) - Applying or detecting watermarks on media - Analyzing media for speaker info, emotion, transcription, or misinformation - Asking natural-language questions about detection results - Matching or verifying speaker identity against known voice profiles - Detecting AI-generated or machine-written text - Any mention of: "deepfake", "fake detection", "synthetic media", "voice verification", "watermark", "media forensics", "authenticity check", "source tracing", "is this real", "AI-written text", "text detection" **Do NOT use** for text-to-speech generation, voice cloning, or speech-to-text transcription — those are separate Resemble capabilities. ## Capability Decision Tree | User wants to... | Use this | API endpoint | |-------------------------------------------------------|---------------------------|----------------------------| | Check if media is AI-generated / deepfake | **Deepfake Detection** | `POST /detect` | | Know *which AI platform* made fake audio | **Audio Source Tracing** | `POST /detect` with flag | | Get speaker info, emotion, transcription from media | **Intelligence** | `POST /intelligence` | | Ask questions about a completed detection | **Detect Intelligence** | `POST /detects/{uuid}/intelligence` | | Apply an invisible watermark to media | **Watermark Apply** | `POST /watermark/apply` | | Check if media contains a watermark | **Watermark Detect** | `POST /watermark/detect` | | Verify a speaker's identity against known profiles | **Identity Search** | `POST /identity/search` | | Check if text is AI-generated | **Text Detection** | `POST /text_detect` | | Create a voice identity profile for future matching | **Identity Create** | `POST /identity` | When multiple capabilities apply (e.g., user wants deepfake detection AND intelligence), combine them in a single `POST /detect` call using the `intelligence: true` flag rather than making separate requests. ## Required Setup - **API Key**: Bearer token from the Resemble AI dashboard - **Base URL**: `https://app.resemble.ai/api/v2` - **Auth Header**: `Authorization: Bearer ` - **Media Requirement**: All media must be at a publicly accessible HTTPS URL If the user provides a local file path instead of a URL, inform them the file must be hosted at a public HTTPS URL first. Do not attempt to upload local files to the API. ## MCP Tools Available When the Resemble MCP server is connected, use these tools instead of raw API calls: | Tool | Purpose | |---------------------------|---------------------------------------------------| | `resemble_docs_lookup` | Get comprehensive docs for any detect sub-topic | | `resemble_search` | Search across all documentation | | `resemble_api_endpoint` | Get exact OpenAPI spec for any endpoint | | `resemble_api_search` | Find endpoints by keyword | | `resemble_get_page` | Read specific documentation pages | | `resemble_list_topics` | List all available topics | **Tool usage pattern**: Use `resemble_docs_lookup` with topic `"detect"` to get the full picture, then `resemble_api_endpoint` for exact request/response schemas before making API calls. --- ## Phase 1: Deepfake Detection The core capability. Submit any audio, image, or video for AI-generated content analysis. ### Submit a Detection ``` POST /detect Content-Type: application/json Authorization: Bearer { "url": "https://example.com/media.mp4", "visualize": true, "intelligence": true, "audio_source_tracing": true } ``` **Parameters:** | Parameter | Type | Required | Description | |------------------------|---------|----------|----------------------------------------------------------| | `url` | string | Yes | HTTPS URL to audio, image, or video file | | `callback_url` | string | No | Webhook URL for async completion notification | | `visualize` | boolean | No | Generate heatmap/visualization artifacts | | `intelligence` | boolean | No | Run multimodal intelligence analysis alongside detection | | `audio_source_tracing` | boolean | No | Identify which AI platform synthesized fake audio | | `frame_length` | integer | No | Audio/video analysis window size in seconds (1–4, default 2) | | `start_region` | number | No | Start of segment to analyze (seconds) | | `end_region` | number | No | End of segment to analyze (seconds) | | `model_types` | string | No | `"image"` or `"talking_head"` (for face-swap detection) | | `use_reverse_search` | boolean | No | Enable reverse image search (image only) | | `use_ood_detector` | boolean | No | Enable out-of-distribution detection | | `zero_retention_mode` | boolean | No | Auto-delete media after detection completes | **Supported formats:** - Audio: WAV, MP3, OGG, M4A, FLAC - Video: MP4, MOV, AVI, WMV - Image: JPG, PNG, GIF, WEBP ### Poll for Results Detection is asynchronous. Poll `GET /detect/{uuid}` until `status` is `"completed"` or `"failed"`. ``` GET /detect/{uuid} Authorization: Bearer ``` **Polling best practice:** Start at 2s intervals, back off to 5s, then 10s. Most detections complete within 10–60 seconds depending on media length. ### Reading Results by Media Type **Audio results** — in `metrics`: ```json { "label": "fake", "score": ["0.92", "0.88", "0.95"], "consistency": "0.91", "aggregated_score": "0.92", "image": "https://..." } ``` - `label`: `"fake"` or `"real"` — the verdict - `score`: Per-chunk prediction scores (array) - `aggregated_score`: Overall confidence (0.0–1.0, higher = more likely synthetic) - `consistency`: How consistent the prediction is across chunks - `image`: Visualization heatmap URL (if `visualize: true`) **Image results** — in `image_metrics`: ```json { "type": "ImageAnalysis", "label": "fake", "score": 0.87, "image": "https://...", "ifl": { "score": 0.82, "heatmap": "https://..." }, "reverse_image_search_sources": [ { "url": "...", "title": "...", "verdict": "known_fake", "similarity": 0.95 } ] } ``` - `label` / `score`: Verdict and confidence - `ifl`: Invisible Frequency Layer analysis with heatmap - `reverse_image_search_sources`: Known sources found online (if `use_reverse_search: true`) **Video results** — in `video_metrics`: ```json { "label": "fake", "score": 0.89, "certainty": 0.91, "children": [ { "type": "VideoResult", "conclusion": "Fake", "score": 0.89, "timestamp": 2.5, "children": [...] } ] } ``` - Hierarchical tree of frame-level and segment-level results - Each child has `timestamp`, `score`, `certainty`, and may have nested `children` - Video with audio track returns both `metrics` (audio) and `video_metrics` (visual) ### Interpreting Scores | Score Range | Interpretation | |-------------|-----------------------------------------------------| | 0.0 – 0.3 | Strong indication of authentic/real media | | 0.3 – 0.5 | Inconclusive — recommend additional analysis | | 0.5 – 0.7 | Likely synthetic — flag for review | | 0.7 – 1.0 | High confidence synthetic/AI-generated | **Always present scores with context.** Say "The detection returned a score of 0.87, indicating high confidence that this audio is AI-generated" — never just "it's fake." --- ## Phase 2: Intelligence — Media Analysis Analyze media for rich structured insights independent of or alongside detection. ### Standalone Intelligence ``` POST /intelligence Content-Type: application/json Authorization: Bearer { "url": "https://example.com/audio.mp3", "json": true } ``` **Parameters:** | Parameter | Type | Required | Description | |----------------|---------|----------|----------------------------------------------------------| | `url` | string | One of | HTTPS URL to media file | | `media_token` | string | One of | Token from secure upload (alternative to URL) | | `detect_id` | string | No | UUID of existing detect to associate | | `media_type` | string | No | `"audio"`, `"video"`, or `"image"` (auto-detected) | | `json` | boolean | No | Return structured fields (default: false for audio/video, true for image) | | `callback_url` | string | No | Webhook for async mode | **Audio/Video structured response** (`json: true`): - `speaker_info` — speaker description (age, gender) - `language` / `dialect` — detected language - `emotion` — detected emotional state - `speaking_style` — conversational, formal, etc. - `context` — inferred context of the speech - `message` — content summary - `abnormalities` — anomalies detected in the media - `transcription` — full transcript - `translation` — translation if non-English - `misinformation` — misinformation analysis **Image structured response:** - `scene_description` — what the image shows - `subjects` — people/objects identified - `authenticity_analysis` — visual authenticity assessment - `context_and_setting` — environment description - `abnormalities` — visual anomalies - `misinformation` — misinformation analysis ### Detect Intelligence — Ask Questions About Results After a detection completes, ask natural-language questions about it: ``` POST /detects/{detect_uuid}/intelligence Content-Type: application/json Authorization: Bearer { "query": "How confident is the model that this audio is fake?" } ``` This returns a question UUID. Poll `GET /detects/{detect_uuid}/intelligence/{question_uuid}` until `status` is `"completed"` to get the `answer`. **Good questions to suggest:** - "Summarize the detection results in plain language" - "What specific indicators suggest this is AI-generated?" - "How do the audio and video detection results differ?" - "What is the confidence level and what does it mean?" - "Are there any inconsistencies in the analysis?" **Status flow:** `pending` → `processing` → `completed` (or `failed`) **Prerequisite:** The detection must have `status: "completed"`. Submitting a question against a processing or failed detection returns a 422 error. --- ## Phase 3: Audio Source Tracing When audio is detected as synthetic (`label: "fake"`), identify which AI platform generated it. **Enable it** by setting `audio_source_tracing: true` in the `POST /detect` request. **Result** appears in the detection response under `audio_source_tracing`: ```json { "label": "elevenlabs", "error_message": null } ``` Known source labels include: `resemble_ai`, `elevenlabs`, `real`, and others as the model expands. **Important:** Source tracing only runs when audio is labeled as `"fake"`. If the audio is `"real"`, no source tracing result will appear. **Standalone query:** - `GET /audio_source_tracings` — list all source tracing reports - `GET /audio_source_tracings/{uuid}` — get specific report --- ## Phase 4: Watermarking Apply invisible watermarks to media for provenance tracking, or detect existing watermarks. ### Apply a Watermark ``` POST /watermark/apply Content-Type: application/json Authorization: Bearer Prefer: wait { "url": "https://example.com/image.png", "strength": 0.3, "custom_message": "my-organization" } ``` | Parameter | Type | Required | Description | |------------------|--------|----------|------------------------------------------------------| | `url` | string | Yes | HTTPS URL to media file | | `strength` | number | No | Watermark strength 0.0–1.0 (image/video only, default 0.2) | | `custom_message` | string | No | Custom message to embed (image/video only, default "resembleai") | - Add `Prefer: wait` header for synchronous response - Without it, poll `GET /watermark/apply/{uuid}/result` - Response includes `watermarked_media` URL to download the watermarked file ### Detect a Watermark ``` POST /watermark/detect Content-Type: application/json Authorization: Bearer Prefer: wait { "url": "https://example.com/suspect-image.png" } ``` **Audio detection result:** ```json { "has_watermark": true, "confidence": 0.95 } ``` **Image/Video detection result:** ```json { "has_watermark": true } ``` --- ## Phase 5: Identity — Speaker Verification (Beta) Create voice identity profiles and match incoming audio against them. > **Beta feature** — requires joining the preview program. Inform the user if they encounter access errors. ### Create an Identity Profile ``` POST /identity Content-Type: application/json Authorization: Bearer { "audio_url": "https://example.com/known-speaker.wav", "name": "Jane Doe" } ``` ### Search Against Known Identities ``` POST /identity/search Content-Type: application/json Authorization: Bearer { "audio_url": "https://example.com/unknown-speaker.wav", "top_k": 5 } ``` **Response:** ```json { "success": true, "item": [ { "uuid": "...", "name": "Jane Doe", "confidence": 0.92, "distance": 0.08 } ] } ``` Lower `distance` = closer match. Higher `confidence` = stronger match. --- ## Phase 6: Text Detection Detect whether text content is AI-generated or human-written. > **Beta feature** — requires the `detect_beta_user` role or a billing plan that includes the `dfd_text` product. ### Submit a Text Detection ``` POST /text_detect Content-Type: application/json Authorization: Bearer ``` Add the `Prefer: wait` header for a synchronous (blocking) response. Without it, the job runs asynchronously — poll or use a callback. **Parameters:** | Parameter | Type | Required | Description | |----------------|---------|----------|----------------------------------------------------------| | `text` | string | Yes | Text to analyze (max 100,000 characters) | | `thinking` | string | No | Always use `"low"` (default) | | `threshold` | float | No | Decision threshold 0.0–1.0 (default: 0.5) | | `callback_url` | string | No | Webhook URL for async completion notification | | `privacy_mode` | boolean | No | If true, text content is not stored after analysis | **Response:** ```json { "success": true, "item": { "uuid": "abc-123", "status": "completed", "prediction": "ai", "confidence": 0.91, "text_content": "This is some text to analyze.", "privacy_mode": false, "created_at": "...", "updated_at": "..." } } ``` - `prediction`: `"ai"` or `"human"` — the verdict - `confidence`: 0.0–1.0, higher = more confident in the prediction - `status`: `"processing"`, `"completed"`, or `"failed"` ### Poll for Results If you did not use `Prefer: wait`, poll until `status` is `"completed"` or `"failed"`: ``` GET /text_detect/{uuid} Authorization: Bearer ``` ### List Text Detections ``` GET /text_detect Authorization: Bearer ``` Returns paginated text detections for the team. ### Callback If `callback_url` was provided, a `POST` is sent on completion: ```json { "success": true, "item": { ... } } ``` On failure: ```json { "success": false, "item": { ... }, "error": "Error message here" } ``` --- ## Recommended Workflows ### Full Media Forensics (Most Thorough) For a comprehensive analysis, combine all capabilities: 1. Submit detection with all flags enabled: ```json { "url": "https://example.com/suspect.mp4", "visualize": true, "intelligence": true, "audio_source_tracing": true, "use_reverse_search": true } ``` 2. Poll until `status: "completed"` 3. Read `metrics` / `image_metrics` / `video_metrics` for the verdict 4. Read `intelligence.description` for structured media analysis 5. If audio labeled `"fake"`, check `audio_source_tracing.label` for the source platform 6. Ask follow-up questions via Detect Intelligence if anything needs clarification 7. Check for watermarks via `POST /watermark/detect` if provenance is relevant ### Quick Authenticity Check (Fastest) For a fast pass/fail: 1. Submit minimal detection: `{ "url": "..." }` 2. Poll until complete 3. Check `label` and `aggregated_score` (audio) or `label` and `score` (image/video) 4. Report result with score context ### Provenance Pipeline (Content Creators) For creators who want to prove their content is authentic: 1. Apply watermark to original content: `POST /watermark/apply` 2. Distribute watermarked media 3. Later, verify provenance: `POST /watermark/detect` against any copy --- ## Red Flags — Stop and Reassess - **Declaring authenticity without a detection result** — Never say media is real or fake based on visual/auditory inspection alone - **Ignoring the score and reporting only the label** — A `"fake"` label with score 0.51 means something very different from score 0.95 - **Submitting local file paths to the API** — The API requires publicly accessible HTTPS URLs (does not apply to text detection) - **Sending text longer than 100,000 characters to text detection** — Split into chunks or inform the user of the limit - **Polling too aggressively** — Start at 2s intervals, back off exponentially; do not loop at <1s - **Asking Detect Intelligence questions before detection completes** — Results in 422 error - **Expecting source tracing on "real" audio** — Source tracing only runs on audio labeled `"fake"` - **Treating beta features (Identity) as production-ready** — Warn users about beta status - **Ignoring `zero_retention_mode` for sensitive media** — Always suggest this flag when the user indicates the media is sensitive or private - **Making multiple separate API calls when flags can combine** — Use `intelligence: true` and `audio_source_tracing: true` on the detection call instead of separate requests ## Response Presentation Guidelines When presenting results to users: 1. **Lead with the verdict** — "The detection indicates this audio is likely AI-generated (score: 0.87)" 2. **Provide score context** — Use the score interpretation table above 3. **Mention limitations** — Detection is probabilistic, not absolute proof 4. **Include actionable next steps** — Suggest intelligence queries, source tracing, or watermark checks as appropriate 5. **For inconclusive results (0.3–0.5)** — Explicitly state the result is inconclusive and recommend additional analysis with different parameters or manual review 6. **Never present detection as legal evidence** — Detection results are analytical tools, not forensic certifications ## Error Handling | Error | Cause | Resolution | |-----------|------------------------------------------|--------------------------------------------------| | 400 | Invalid request body or missing `url` | Check required parameters | | 401 | Invalid or missing API key | Verify `RESEMBLE_API_KEY` | | 404 | Detection UUID not found | Verify the UUID from the creation response | | 422 | Detection not completed (for Intelligence) | Wait for detection to reach `completed` status | | 429 | Rate limited | Back off and retry with exponential delay | | 500 | Server error | Retry once, then report to user | ## Privacy & Compliance Notes - **Zero retention mode**: Set `zero_retention_mode: true` to auto-delete media after analysis. The URL is redacted and `media_deleted` is set to true post-completion. - **Text privacy mode**: Set `privacy_mode: true` on text detection to prevent text content from being stored after analysis. - **Data handling**: Media URLs and text content are stored by default. For GDPR/compliance-sensitive workflows, enable zero retention (media) or privacy mode (text). - **Callback security**: If using `callback_url`, ensure the endpoint is HTTPS and authenticated on the receiving end.