Tip: For maximum performance, we recommend storing source video files on an SSD drive.

• Trial period: First 3 days free. If you plan to continue using the program, you need to pay for a subscription in the bot.
• Device limit: One subscription = 1 device.
• Additional slots: Want to work on both PC and laptop? You can purchase additional device slots in the bot menu.
• Tokens: The app's AI features — SEO generation in "Packaging", title ideas and thumbnail generation — spend tokens from your balance. The cost of each action is shown right in the interface before you run it. Local Whisper transcription works offline and spends no tokens. You can top up your balance in the Telegram bot or on the website.

Enable this toggle if the wide shot recording is split into multiple files (e.g., 001.MP4, 002.MP4).

Add 1 to 7 speakers. For each one, specify:

• Name: Track name in your editing software.
• Color: Track color on the timeline (Premiere Pro).
• Video: Slot for video files.
  You can load one file or multiple (if recording was stopped or multicam is used).
  To load multiple files, enable the "Multiple files" toggle under the slot.
• Audio: Slot for a separate audio track (optional).
  If the slot is empty, the program will use audio from the video file.
• Audio channel (L / R / Mono): NEW
  Below the "Audio" slot there are two dropdowns:
  • Video source — which file to take the sound from: "Own video", "Wide Shot" or another speaker's video.
  • Channel — which channel to use: Mono (the whole signal), Left (L) or Right (R).
  
  This is useful when a speaker's sound is recorded into the L or R channel of another speaker's video file. Pick the right file in "Source" and select the channel — the program will extract exactly that signal.

Filming yourself on several cameras (blog, expert content)? Add just one speaker — and the program will produce a multicam edit from a single "talker", automatically switching angles and breaking the picture up with the wide shot.

In this mode the program automatically hides the irrelevant settings (interruptions, switch delay, pause between wide shots) — they're only needed for multi-participant dialogues. The interface stays clean.

Final parameters before creating the XML file.

Load a finished video or audio — the program "listens" to it and generates a full SEO package for YouTube. No need to come up with titles and place timecodes manually.

"Packaging" requires local Whisper transcription (enabled by default). Without it analysis is impossible — see the "Settings" section.

1. Choose a Video / Audio file.
2. Pick an AI model and click "Generate".

Extra materials to promote the video in one click:

• Pinned comment — text worth pinning under the video.
• Telegram post — a ready announcement for your channel.
• Community post — for the Community tab on YouTube.
• Shorts ideas — hints on which moments to cut into short vertical clips.

A separate tab with video analytics: a virality estimate, attention retention and editing tips — what to pay attention to next time.

"Packaging" only works with local Whisper transcription enabled — the program analyzes the actual text of your speech. Whisper is on by default and installs automatically; you can manage it in the "Settings" section. The more accurate the model, the more accurate the timecodes (down to ±5 sec).

1. Enter the title that will be on the thumbnail.
   The "Title ideas" button suggests variants based on the analysis of your video from the "Packaging" tab.
2. Describe the background: mood, emotions, colors, objects.
3. Add the speaker's face — upload a photo or grab a frame straight from the video.
4. Choose the aspect ratio and quality, then click "Generate".

Each thumbnail generation spends tokens. The cost depends on the chosen model and quality (1K / 2K / 4K) — the exact number is shown right on the button before you run it.

Local audio transcription right on your computer (offline, nothing sent to the cloud). Enabled by default and required for "Packaging" — without it, podcast analysis is impossible. Dramatically improves chapter timecode accuracy — down to ±5 seconds instead of several minutes.

Models to choose from

Dependencies (mlx-whisper for Apple Silicon or faster-whisper for Intel/Windows) install automatically. The model is downloaded once and cached on your device — in this same section you can manage the cache and delete unneeded models.

Advanced mode for the "Editing" tab: a neural network tells speakers apart by voice. A free HuggingFace token is required — connection instructions are right inside the app.

Here you set the preview generation parameters for the "Thumbnails" tab: