Simple AI Video Clipper - Generate Shorts, Viral title, Hashtag, Download video clips for social media from long videos

Automatically detect the most engaging moments in videos (direct video files or long YouTube videos such as podcasts, interviews, and streams) and generate ready-to-use clips optimized for YouTube Shorts, Instagram Reels, and TikTok. Produce Title, hastag and short description for clips.

High-Level Workflow

The Actor runs an automated pipeline to ingest, analyze, and produce ready-to-use social clips:

1. Input Validation: Accepts either a direct video file (file_video) or a YouTube link (youtube_url) under strict mutual exclusivity.
2. Audio & Speech Extraction: Ingests audio and subtitle streams (via embedded tracks, speech-to-text, or YouTube caption streams) to obtain precise word-level timestamps.
3. Salient Moment Detection: Analyzes dialogue flow, retention spikes, and pacing to identify high-potential viral highlights.
4. Context Framing & Subtitles: Attaches balanced pre-roll and post-roll context buffers around each highlight and generates dynamic animated karaoke subtitles.
5. Output Generation: Trims downloadable MP4 video clips (for file_video) or produces direct timestamp links (for youtube_url), accompanied by complete social media copy kits (viral_hook, Caption, Hashtag, Virality_score_description, transcript) and ready-to-use FFmpeg commands.

A visual overview of the end-to-end processing pipeline is shown below:

---

Input Parameters

Video Input Source Exclusivity Rule

The file_video and youtube_url parameters are mutually exclusive:

• Exactly one of the two parameters must be provided. If both or neither are provided, the Actor stops execution and raises a validation error.
• When file_video is used: The video file is downloaded, processed, and automatically cut into clips. You get the pre-cut video file ready for direct download via the download_url field, while Download Link (download_link) is set to null (null). A ready-to-use FFmpeg command to extract and encode the clip is also provided.
• When youtube_url is used: The video file is NOT downloaded. Instead, you get direct links to the exact timestamps of the detected clips via Download Link (download_link, direct_timestamp_url), while download_url remains empty/null (null). A ready-to-use FFmpeg command is also provided if you wish to extract and encode the clip locally.
• Ready-to-use Social Media Copy Kit (included in both modes): Regardless of which input option is used (file_video or youtube_url), every output record always provides complete social media data (viral_hook, Caption, Hashtag, Virality_score_description, and transcript) ready to be copy-pasted directly to social networks (Instagram Reels, TikTok, YouTube Shorts).

Input Parameters Table

Parameter	Type	Required	Default	Description
file_video	String	No*	—	Direct URL to a video file (.mp4, .mov, .mkv, .webm). Mutually exclusive with youtube_url. When used, the video is downloaded and trimmed, giving you the pre-cut video file ready for download in download_url.
youtube_url	String	No*	—	URL of a YouTube video (e.g., https://www.youtube.com/watch?v=... or https://youtu.be/...). Mutually exclusive with file_video. The video is not downloaded; instead, you get the direct link to the exact clip timestamp in Download Link.
max_clips	Integer	No	5	Maximum number of top highlight clips to extract (1 to 50).
max_length	Integer	No	60	Maximum duration in seconds for each clip (10 to 300 seconds).
targetFormat	String	No	INSTAGRAM_REELS_VERTICAL_9_16_HIGHRES_1080P_MP4	Target format, aspect ratio, and resolution for social media clips. Supports vertical 9:16 profiles (1080p and 720p for Reels, TikTok, Shorts), square 1:1, portrait 4:5, or original format.
burn_subtitles	Boolean	No	true	If enabled (true), burns dynamic animated karaoke-style subtitles (with active word-by-word highlight) directly onto the video.
subtitle_style	String	No	HORMOZI_YELLOW	Visual typography preset for animated subtitles (HORMOZI_YELLOW, MRBEAST_BOUNCE, CLEAN_MINIMAL, NEON_CYAN).
subtitle_position	String	No	CENTER	Vertical subtitle position on screen (BOTTOM, CENTER, TOP). CENTER is placed in the social safe zone above native UI elements.
burn_hook_header	Boolean	No	false	When enabled, overlays a prominent high-contrast headline banner showing the viral hook during the opening 4 seconds.
focus_topic	String	No	—	Optional theme or keyword filter (e.g. "artificial intelligence", "investing"). Directs AI to prioritize moments discussing this topic.
normalize_audio	Boolean	No	false	When enabled, normalizes audio to -14 LUFS (EBU R128 broadcast standard) for loud, clear mobile speech.
generate_thumbnails	Boolean	No	false	When enabled, extracts a high-resolution JPEG cover frame from each clip and provides a public link in thumbnail_url.
layout_mode	String	No	AUTO_CENTER_CROP	Social video composition layout (AUTO_CENTER_CROP, PODCAST_SPLIT_STACK, BLURRED_BARS).
smart_face_tracking	Boolean	No	false	Automatically detects speaker faces in video frames to center the vertical crop on the active speaker.
webhook_url	String	No	—	Optional HTTP or HTTPS webhook URL. Sends a JSON POST payload with all clip data and download URLs upon run completion.

* Exactly one of file_video or youtube_url must be specified.

Example Input: Direct Video File with Creator Enhancements

{
  "file_video": "https://github.com/ilborso/resources/raw/refs/heads/main/Interstella.mp4",
  "max_clips": 3,
  "max_length": 30,
  "targetFormat": "INSTAGRAM_REELS_VERTICAL_9_16_HIGHRES_1080P_MP4",
  "burn_subtitles": true,
  "subtitle_style": "HORMOZI_YELLOW",
  "subtitle_position": "CENTER",
  "burn_hook_header": true,
  "focus_topic": "watchmaking",
  "normalize_audio": true,
  "generate_thumbnails": true,
  "layout_mode": "AUTO_CENTER_CROP",
  "smart_face_tracking": true
}

Example Input: YouTube Video Link

{
  "youtube_url": "https://www.youtube.com/watch?v=jfKfPfyJRdk",
  "max_clips": 5,
  "max_length": 60,
  "targetFormat": "TIKTOK_FEED_VERTICAL_9_16_HIGHRES_1080P_MP4",
  "burn_subtitles": true,
  "subtitle_style": "MRBEAST_BOUNCE",
  "subtitle_position": "CENTER",
  "burn_hook_header": true,
  "webhook_url": "https://hook.eu1.make.com/your-custom-webhook-endpoint"
}

---

Pre-uploading Local Videos to Apify Key-Value Store

Use Case

If your video file is stored locally on your machine or private server rather than hosted on a public URL (such as YouTube, GitHub, or AWS S3), you can upload it directly to an Apify Key-Value Store before running the Actor.

Why use this approach?

• Bypasses Input Size Limits: Apify Actor input payloads are JSON documents limited to approximately 9 MB, making inline Base64 video encoding impractical. Uploading to a Key-Value Store supports large video files (hundreds of megabytes or gigabytes).
• Zero Base64 Overhead: Avoids the 33% payload size inflation inherent to Base64 encoding.
• High-Speed Cloud Streaming: The Actor downloads and processes the video directly within the Apify cloud infrastructure with high bandwidth and minimal latency.

Once uploaded, pass the Key-Value Store record URL to the file_video parameter to trigger processing.

Upload via cURL

Upload your local video file directly to an Apify Key-Value Store record using a binary PUT request:

curl -X PUT "https://api.apify.com/v2/key-value-stores/<STORE_ID_OR_NAME>/records/<RECORD_KEY>?token=<YOUR_APIFY_TOKEN>" \
  -H "Content-Type: video/mp4" \
  --data-binary "@my_local_video.mp4"

Upload via Windows Powershell

Upload your local video file directly to an Apify Key-Value Store record using Invoke-RestMethod

$uri = "https://api.apify.com/v2/key-value-stores/<STORE_ID_OR_NAME>/records/<RECORD_KEY>?token=<YOUR_APIFY_TOKEN>"
$filePath = (Resolve-Path ".\my_local_video.mp4").Path

Invoke-RestMethod -Uri $uri -Method Put -ContentType "video/mp4" -InFile $filePath

Once uploaded, obtain the direct record URL:

https://api.apify.com/v2/key-value-stores/<STORE_ID_OR_NAME>/records/<RECORD_KEY>

Pass this URL to the file_video input parameter when running the Actor.

---

Output Dataset Structure

Processed results are saved to the default Apify dataset with the following structure. In both modes (file_video and youtube_url), every record provides complete, ready-to-publish social media data (viral_hook, Caption, Hashtag, Virality_score_description, and transcript):

Output Fields Table

Field	Type	Description
rank	Integer	Rank position of the highlight moment (1 = highest score).
Title	String	Catchy title generated for the clip (pure text, no icons or emojis).
score	Float	Composite engagement and virality score from 0 to 100.
viral_hook	String	Opening 3-second hook for on-screen attention (pure text, no icons or emojis).
Caption	String	Social media post copy with narrative context and call-to-action (pure text, no icons or emojis).
Hashtag	Array	List of 5-8 relevant and trending hashtags for social posting.
Virality_score_description	String	Explanation of the virality score rationale and retention triggers (pure text, no icons or emojis).
start_time	String	Clip start timestamp formatted as HH:MM:SS.
end_time	String	Clip end timestamp formatted as HH:MM:SS.
duration_seconds	Float	Clip duration in seconds.
target_format	String	Video format and resolution profile applied to the clip.
detection_reasons	Array	Signals and factors that triggered the selection of this highlight.
transcript	String	Spoken dialogue transcribed within the highlight segment.
download_url	String / Null	Direct download link for the cut MP4 video clip. Populated when input is file_video; null (null) when input is youtube_url.
Download Link (download_link)	String / Null	Direct YouTube link jumping to the highlight timestamp. Populated when input is youtube_url; null (null) when input is file_video.
direct_timestamp_url	String / Null	Direct YouTube link jumping to the start of the clip (returned when youtube_url is used).
clip_filename	String / Null	Filename of the trimmed clip (e.g. clip_1.mp4, returned when file_video is used).
subtitles_url	String / Null	Download link for the animated ASS subtitle file.
thumbnail_url	String / Null	Direct download link for the extracted high-resolution cover JPEG thumbnail (populated when file_video is used and generate_thumbnails is enabled).
ffmpeg_command	String	Ready-to-use FFmpeg command to extract and encode the clip.

Example Output Record (when using file_video)

{
  "rank": 1,
  "Title": "The Interstellar Watch You Can Actually Buy",
  "score": 92.0,
  "start_time": "00:00:00",
  "end_time": "00:00:25",
  "duration_seconds": 25.0,
  "target_format": "YOUTUBE_SHORTS_VERTICAL_9_16_LOWRES_720P_MP4",
  "viral_hook": "The Interstellar Movie Watch",
  "Caption": "Did you know you can buy the exact watch from Christopher Nolan's Interstellar? Here is everything you need to know about the Hamilton Khaki Murph 42mm. Would you wear this everyday? Let us know in the comments.",
  "Hashtag": [
    "#shorts",
    "#reels",
    "#viral",
    "#interstellar",
    "#hamiltonwatch",
    "#watchcollecting"
  ],
  "Virality_score_description": "Leverages mass pop culture appeal from a beloved movie like Interstellar, instantly capturing viewer attention within the first second.",
  "detection_reasons": [
    "The Interstellar Watch You Can Actually Buy",
    "Mentions a major blockbuster movie connection immediately",
    "+10s_context_buffer"
  ],
  "transcript": "A month ago, I bought a Hamilton Carkey Murph 42mm as a bit of a treat for myself and I've worn it every single day since...",
  "clip_filename": "clip_1.mp4",
  "clip_size_mb": 1.24,
  "download_url": "https://api.apify.com/v2/key-value-stores/1DMJuZ8OdjTGq8YZW/records/clip_1.mp4?signature=tJvlQNgO8lanHr7Rv1Ua",
  "download_link": null,
  "Download Link": null,
  "subtitles_url": "https://api.apify.com/v2/key-value-stores/1DMJuZ8OdjTGq8YZW/records/clip_1.ass?signature=O0Olrqd0uWio81s4dkkV",
  "thumbnail_url": "https://api.apify.com/v2/key-value-stores/1DMJuZ8OdjTGq8YZW/records/clip_1_thumb.jpg?signature=xY82NzlaK892",
  "ffmpeg_command": "ffmpeg -ss 00:00:00 -to 00:00:25 -i \"https://github.com/ilborso/resources/raw/refs/heads/main/Interstella.mp4\" -vf \"scale=720:1280:force_original_aspect_ratio=increase,crop=720:1280:'trunc((in_w-720)/4)*2':'trunc((in_h-1280)/4)*2',setsar=1,ass='clip_1.ass'\" -c:v libx264 -pix_fmt yuv420p -preset fast -crf 22 -c:a aac -b:a 128k -movflags +faststart clip_1.mp4"
}

Example Output Record (when using youtube_url)

{
  "rank": 1,
  "Title": "Epic Game Winning Play",
  "score": 92.4,
  "start_time": "01:23:45",
  "end_time": "01:24:30",
  "duration_seconds": 45.0,
  "target_format": "TIKTOK_FEED_VERTICAL_9_16_HIGHRES_1080P_MP4",
  "viral_hook": "You will not believe this final play...",
  "Caption": "Down to the final 3 seconds! Watch what happens next. Drop your reaction in the comments!",
  "Hashtag": [
    "#shorts",
    "#reels",
    "#tiktok",
    "#viral",
    "#fyp"
  ],
  "Virality_score_description": "High emotional volume spike with strong climax payoff, rapid chat velocity, and instant comment debate potential.",
  "detection_reasons": [
    "audio_volume_spike_+180%",
    "chat_velocity_320%",
    "+10s_context_buffer"
  ],
  "transcript": "Oh my god, did you see that?! That was absolutely incredible! I can't believe that just happened!",
  "download_url": null,
  "download_link": "https://youtu.be/jfKfPfyJRdk?t=5025",
  "Download Link": "https://youtu.be/jfKfPfyJRdk?t=5025",
  "direct_timestamp_url": "https://youtu.be/jfKfPfyJRdk?t=5025",
  "thumbnail_url": null,
  "ffmpeg_command": "ffmpeg -ss 01:23:45 -to 01:24:30 -i \"$(yt-dlp -g -f 'bestvideo[ext=mp4]+bestaudio[ext=m4a]/best[ext=mp4]/best' 'https://www.youtube.com/watch?v=jfKfPfyJRdk')\" -vf \"scale=1080:1920:force_original_aspect_ratio=increase,crop=1080:1920:'trunc((in_w-1080)/4)*2':'trunc((in_h-1920)/4)*2',setsar=1,ass='clip_1.ass'\" -c:v libx264 -pix_fmt yuv420p -preset fast -crf 22 -c:a aac -b:a 128k -movflags +faststart clip_1.mp4"
}

---

Creator and Social Media Enhancements

This Actor includes dedicated features designed specifically to streamline the production workflow of content creators, social media managers, and podcast editors.

1. Animated Subtitle Styles and Visual Typography (subtitle_style)

Captions are critical for short-form retention since a significant portion of social media users browse with audio disabled. Four typography profiles are available:

• HORMOZI_YELLOW (Default): High-visibility bold white text with bright yellow highlights on the active spoken word, surrounded by a heavy black outline and drop shadow.
• MRBEAST_BOUNCE: Dynamic green highlight featuring a subtle scale-up pop animation on each spoken word as it is delivered.
• CLEAN_MINIMAL: Elegant, modern light typography with refined outlines and soft golden highlights, suited for business, luxury, and educational clips.
• NEON_CYAN: Electric cyan glow with dark violet backing, tailored for gaming highlights, tech reviews, and nightlife content.

2. Vertical Subtitle Placement and Safe-Zone Alignment (subtitle_position)

To avoid overlapping native platform interface elements (such as TikTok captions, username links, like buttons, and audio wheels):

• CENTER (Default): Positions captions in the central safe-zone of vertical video formats.
• BOTTOM: Aligns text in the lower third with safe margin buffers.
• TOP: Displays captions across the upper third of the video.

3. Viral Hook Headline Header Bar (burn_hook_header)

When burn_hook_header is set to true, a prominent, high-contrast headline banner displaying the AI-generated viral_hook is burned into the top center of the video during the first 4 seconds of playback. This grabs the viewer's attention immediately within the critical 3-second scroll window before disappearing cleanly into the body of the clip.

4. Topic and Keyword AI Guidance (focus_topic)

When processing long interviews, conferences, or multi-topic podcasts, you can specify a keyword or thematic focus (for example: "real estate", "AI agents", or "funny stories"). The AI prioritizes and scores moments that specifically discuss the requested topic, filtering out irrelevant segments.

5. Mobile Audio Normalization (normalize_audio)

Different recording sources, microphone distances, and podcast guests often produce uneven volume levels. Setting normalize_audio to true applies an EBU R128 two-pass loudness filter (-af loudnorm=I=-14:LRA=7:TP=-1.5) targeting -14 LUFS integrated loudness. This delivers consistent, broadcast-ready clarity optimized for smartphone speakers.

6. High-Resolution Cover Frame Extraction (generate_thumbnails)

When enabled (true), the Actor automatically extracts a sharp JPEG frame from the central point of the highlight clip and uploads it to the Apify Key-Value Store. The resulting thumbnail_url can be used directly as a video cover image in automated social publishing pipelines.

7. Framing Layout and Composition Modes (layout_mode)

When converting horizontal (16:9) footage into vertical (9:16) formats for Reels, TikTok, and Shorts, three composition layouts are supported:

• AUTO_CENTER_CROP (Default): Crops horizontally to the center of the frame or to the detected speaker.
• PODCAST_SPLIT_STACK: Vertically stacks two views (top and bottom half) of a two-person interview or podcast, showing both speaker and host simultaneously without awkward horizontal panning.
• BLURRED_BARS: Preserves the complete original horizontal video centered in the frame, while blurring and scaling the background to fill the 9:16 vertical canvas.

8. Smart Face Tracking and Dynamic Auto-Reframe (smart_face_tracking)

When set to true, the Actor analyzes video frames using computer vision face detection to locate the primary speaker's horizontal position. Instead of blindly cropping the static center, the 9:16 crop window dynamically offsets towards the active speaker, keeping them perfectly framed.

9. Webhook Dispatch for Automated Pipelines (webhook_url)

Connect the Actor directly to automation platforms such as Zapier, Make.com, n8n, or custom API endpoints. By supplying an HTTP or HTTPS URL to webhook_url, the Actor automatically sends a POST request containing a complete JSON payload (event, status, video_title, clips_count, and the full array of results with all download URLs and social media copy) immediately upon completion.