XplicitRecAI Wiki

Overview

XplicitRecAI is a desktop application that records live cam shows from Chaturbate and Stripchat and automatically edits the recordings down to the highlights using AI detection.

The big picture

The application is built around two main modules that form a pipeline:

1. Recording Module — monitors your list of performers, detects when they go online, and records their live streams to disk. Recordings are saved per performer as timestamped video files (mostly .mp4) inside the Video Output Directory.
2. Cutting Module — scans an input directory (usually your recording output), merges recordings that were split, runs AI detection over the video frames, cuts out only the segments with action, and optionally converts the results to MP4.

AI detection classes

The AI models are trained on two detection classes:

• Class 0 — "pussy": explicit visual content.
• Class 1 — "bating": active masturbation.

Both the Recording module (in AI Bandwidth Save mode) and the Cutting module use these detections. With Process Only Specific Content enabled, only the "bating" class triggers processing; otherwise both classes count.

Editions

• Free Edition — recording only, no AI features. Many settings are fixed to safe defaults.
• Paid Edition — full feature set including AI detection, the Cutting module, custom models, and all advanced settings. A 7-day trial starts on first run.

Getting Started

From download to your first automatic highlight clip in a few minutes.

Installation

1. Download the ZIP from the start page — Free Edition (recording only) or Paid Edition (full AI feature set, 7-day trial).
2. Extract the ZIP to C:\XplicitRecAI (recommended).
3. Run XplicitRecAI.exe. The first start may take 30–60 seconds while .NET prepares in the background.
4. Accept the EULA and start your trial when prompted.
5. If the tool does not start, check _log\yyyy-MM-dd_App.log for errors or contact support.

First run

• Make sure your system clock is accurate — recording and merging depend on it.
• Keep plenty of free disk space; the tool shuts down below 1 GB on the primary disk.

First configuration

1. Open Recording Settings and set the Video Output Directory to a drive with lots of space (see File Paths).
2. Decide on Weak Mode: leave it on for weak hardware (recording works without AI), or turn it off to enable AI-driven recording (see AI Detection).
3. Go to the Recording page and add your first performers using their full profile URL — the username alone is not enough (e.g. https://chaturbate.com/performername/ or https://stripchat.com/performername). New performers start in Force Record mode with Max quality — adjust per tile if needed.
4. Paid Edition: open Cutting Settings, point the Video Input Directory at your recording output and the Video Output Directory at your library folder (see File Paths).
5. Do a dry run first — and use a small subset, not your whole collection: if you have a big collection, copy a few performer folders into a separate small test directory and point the input there. Enable Retain Original Videos, process those few files, and check the results before letting the Cutting module loose on the real collection (see Cutting Module).

Updating an existing installation

Updating is fully automatic — no manual downloading, extracting, or file replacement is needed.

1. When a new version is available, an Update button appears in the application.
2. Click it and confirm. The application downloads the latest updater directly from the website, showing a progress bar.
3. The updater starts automatically, closes XplicitRecAI (and Guardian), installs the new version, and you're done — settings, configuration files, and recordings are preserved.

Recording Module

The Recording page is the heart of the application: it manages every performer you want to record, shows their live status, and controls the actual stream capture.

What it does

• Performer management — add performers via their full Chaturbate or Stripchat profile URL (e.g. https://chaturbate.com/performername/). Each performer gets a tile with a live preview image, current room state (online/offline), recording state, and per-performer controls.
• Automatic status monitoring — the app polls the cam site APIs at a configurable interval to detect when a performer goes online, then starts recording automatically (when Autostart Recording is enabled).
• Sequence-based capture — live streams are downloaded as short HLS video sequences. The tile shows live statistics: download speed, current sequence, failed vs. successful sequence writes, FPS, and resolution. If too many sequences fail in a row (default: 8 of the last 12), the recording restarts automatically to stay in sync with the server.
• Per-performer actions — each tile has buttons to open the performer's log file or configuration file, pause/resume recording, and a globe icon to open the performer's room. Middle-clicking the globe opens a custom web search (see CustomWebSearchQueryURL in the settings), which is handy for jumping straight to your Stash library.
• Filtering, sorting & pagination — filter the performer list by text or with hashtag filters: #cb / #sc (by cam site), #rec / #!rec (recording / not recording). Auto-sorting keeps the list ordered, and Items Per Page controls how many tiles are shown.
• Performer ratings — rate performers from 1 to 5 stars directly on their tiles.
• Tor status — a banner in the Recording view shows the live Tor connection state (API requests are routed through Tor by default); if Tor fails to start, the banner points you to the log.

Recording Modes

Every performer has a recording mode that decides when the stream is captured:

Target Quality

Each performer has a target quality. The recorder picks the stream variant closest to your target from what the performer actually broadcasts:

• 480p 30 FPS / 480p 60 FPS
• 720p 30 FPS / 720p 60 FPS
• 1080p 30 FPS / 1080p 60 FPS
• Max — always take the highest quality available (default for new performers).

Output files

Finished recordings are written to the Video Output Directory, one subfolder per performer, with timestamped filenames:

VideoOutputDirectory
├── Performer_01
│   ├── Performer_01-2025-06-18_02-00.mp4
│   └── Performer_01-2025-06-18_04-00.mp4
└── Performer_02
    └── Performer_02-2025-06-18_04-00.mp4

Recordings are downloaded and written as MP4; only older versions of the tool produced .ts files. This structure is exactly what the Cutting module expects as its input, so you can point the cutting input directory at the recording output directory for a fully automatic pipeline.

Filename markers (why a recording ended)

When a recording stops, a short marker is added to the filename that tells you exactly why it ended:

Recording Settings

Everything that controls how streams are captured lives on the Recording Settings page. The tables below list each setting with its default value and what it actually does.

File Paths

Recording Control

Chaturbate & Stripchat Settings

Both cam sites have an identical group of settings, tuned independently:

AI Detection

User Interface & Preview

These toggles control what is shown on each performer tile — username, room state, recording state, quality, FPS, resolution, download speed, sequence statistics (current sequence, last/average download time, successful vs. failed writes), image update timings, and the last room-state check. Turning off what you don't need declutters the tiles and slightly reduces UI load. Additional options:

At the bottom of the page, Show Advanced Settings reveals the options intended for experienced users (hidden by default), and Highlight Advanced Settings visually marks that section so advanced options are easy to distinguish.

Cutting Module

The Cutting module turns raw multi-hour recordings into compact highlight clips. It merges split recordings, scans them with the AI, cuts out only the parts with action, and optionally converts everything to MP4 — fully automatically.

Input requirements

The module works on any folder that follows the performer-subfolder layout (which the Recording module produces automatically):

InputDirectory
├── Performer_01
│   ├── Performer_01-2025-06-18_02-00.mp4
│   └── Performer_01-2025-06-18_04-00.mp4
└── Performer_02
    ├── Performer_02-2025-06-18_04-00.mp4
    └── Performer_02-2025-06-18_07-00.mp4

Supported containers: .mp4, .ts, .mkv, .mov, .avi, .wmv, .flv, .m4v. Only .ts, .mp4, and .mkv are heavily tested and reliable.

The processing pipeline

1. Scan — the input folder is checked on an interval (default: every 180 seconds). Files are only picked up when they are done being written and older than the Minimum File Age (default: 8 hours), so shows that are still running are never processed half-finished.
2. Merge split recordings — recordings that belong to the same show are merged based on file metadata: if the gap between one file's last-modified time and the next file's creation time is within the Video Merge Gap (default: 180 seconds), they are joined. Only files from the same performer subfolder are ever merged. Optionally a visual indicator (loading circle) is inserted where the splits occurred.
3. Rewrite (optional) — original videos can be rewritten to fix missing frames from live recordings, which keeps frame numbering and detection timestamps perfectly aligned.
4. AI detection — the module steps through the video, grabbing a frame every 10–15 seconds (configurable via Post-Detection Skip and No-Detection Skip). Each frame is run through the detection model; frames at or above the confidence threshold (default: 33%) are marked.
5. Clustering — consecutive detections are grouped into events. A new clip starts when no detection occurs for the Detection Event Split Gap (default: 240 seconds). A padding buffer (default: 120 seconds) is added before and after each event so action that starts early or ends late isn't cut off.
6. Cutting — each detection cluster is cut into its own clip. Clips shorter than the Minimum Clip Duration (default: 5 minutes) are discarded. Everything without detections is dropped.
7. Convert & clean up — clips are saved to the output directory, optionally converted to MP4, and the originals are deleted unless you enabled one of the retain options.

Content modes

Via Process Only Specific Content in the Cutting Settings:

• Mode 1 (off) — both detection classes ("pussy" and "bating") produce clips.
• Mode 2 (on) — only "bating" detections produce clips.

Dry runs

To test the process without losing anything, enable Retain Original Videos (and optionally Retain Rewritten Videos). The module then performs one full processing cycle, keeps every source file, and stops. Test on a small subset rather than your whole collection: copy a few performer folders into a separate test directory, point the input there, and check the results before letting it loose on the real collection.

Cutting Settings

Every knob of the cutting pipeline, grouped the same way as on the Cutting Settings page.

File Paths

General Processing

Regular Processing

AI Detection

Timing

The first four timing settings have interactive visualizers directly on the Cutting Settings page, so you can see on a timeline how padding, split gaps, and the two skip intervals interact before committing to values.

Blacklist & Training

The Blacklisted Performers group lets you exclude specific performers from AI processing. Files from blacklisted performers can be moved straight to the output folder without cutting (Process Blacklisted Performer, default: on) — with their own independent copies of the merge/rewrite/retain/MP4 options, so you can, for example, still merge and convert their recordings without any AI cutting. Add usernames to Blacklisted Performer Names to use it.

Like the Recording Settings page, the bottom of the page has Show Advanced Settings to reveal options intended for experienced users, and Highlight Advanced Settings to visually mark that section.

Tips & Related

• GPU acceleration — an NVIDIA GPU with CUDA/cuDNN makes AI detection many times faster for both modules. Follow the GPU Setup guide.
• Recommended pipeline — Recording output directory = Cutting input directory, Cutting output = your library folder, and Stash pointed at the library for browsing. Set the Custom Web Search URL to your Stash instance for one-click lookups from performer tiles.
• Weak Mode — leave it on for weak hardware (recording works fine without AI); turn it off on capable machines to unlock AI Bandwidth Save recording.
• Test first — use the retain options for a dry run on a handful of files before processing a big archive.
• Storage — this is not a lightweight recorder. The tool shuts down if less than 1 GB remains on the primary disk; keep plenty of headroom, and 200+ GB free if you use Local Video Copy.
• Accurate system time — critical for recording and for the metadata-based merging in the Cutting module.
• More detailed step-by-step tutorials will be published here over time. For questions, see the FAQ or contact support.