What Is YT-Short-Clipper? A Beginner-Friendly Guide to jipraks/yt-short-clipper

Repository: https://github.com/jipraks/yt-short-clipper
Topic: automatically turning long YouTube videos into TikTok, Instagram Reels, and YouTube Shorts
Audience: content creators, podcasters, interview channels, vloggers, and developers interested in AI-assisted video editing
Level: beginner-friendly, with minimal jargon
Documentation checked: June 26, 2026

1. What is YT-Short-Clipper in simple words?

YT-Short-Clipper is a desktop application that uses AI to turn a long YouTube video into several short vertical clips.

The basic workflow is:

The repository is mainly designed for content such as:

It is most useful when the source contains clear speech and you want to repurpose it for Shorts, Reels, or TikTok.

2. What problem does it solve?

Manually creating clips from a long video often requires many separate tasks:

YT-Short-Clipper combines these steps into one application.

The realistic value is not that AI produces a perfect final video. Its value is:

3. How does the pipeline work?

Step 1: Download video and subtitles

The application uses yt-dlp to download the source video, retrieve automatically generated subtitles, and extract metadata such as the title, description, and channel name.

The README describes a default maximum download quality of 1080p.

Step 2: Detect highlights with AI

The application parses an SRT subtitle file and sends the timestamped transcript to an AI model.

The model looks for moments such as:

The documented defaults are five clips, each between 60 and 120 seconds, with a target length of about 90 seconds.

An important limitation is:

A visually interesting moment with little or no dialogue may therefore be missed.

Step 3: Cut the selected sections

After AI returns timestamps, FFmpeg cuts the original video into separate segments.

The timestamps can still be imperfect. A clip may start too late, lose necessary context, or end before a sentence is complete. Every result should be reviewed.

Step 4: Convert the clip to portrait

The repository produces 1080 × 1920 output for a 9:16 vertical frame.

It offers two face-tracking modes:

OpenCV: faster and generally crops around the largest detected face.

MediaPipe: attempts to identify the active speaker through lip movement. The README says it may be more accurate in conversations but can be two to three times slower.

Speaker changes use a camera-cut style rather than continuous smooth panning.

Step 5: Generate a hook

The application can build a short introduction containing:

The goal is to tell viewers immediately what the clip is about.

Step 6: Generate captions

The audio is sent to the Whisper API for transcription. The application then creates an ASS subtitle file with word-level timing and burns the captions into the video using FFmpeg.

The styling resembles common short-form captions:

The README explicitly states that the project uses the Whisper API, not a fully local Whisper model.

Step 7: Generate metadata

The application can generate:

A result folder may include a data.json file containing this information.

4. Main features

According to the README, the main features include:

• Automatic video and subtitle download through yt-dlp.
• AI-assisted highlight selection.
• Automatic clipping at selected timestamps.
• Horizontal-to-vertical 9:16 conversion.
• OpenCV and MediaPipe face-tracking modes.
• AI-generated hooks and TTS voice-over.
• Word-by-word captions through the Whisper API.
• Custom watermark support.
• SEO-oriented title and description generation.
• Windows and macOS support.
• GPU encoding through NVENC, AMF, QSV, or VideoToolbox.

5. Is it completely free?

The source code is available under the MIT License and can be downloaded for free. The complete workflow may still cost money.

Potential paid services include:

The README estimates approximately USD 0.10–0.25 per source video for five clips in one sample OpenAI-based configuration. This is only a documentation estimate. Actual pricing can change and depends on the provider, selected models, transcript length, and number of clips.

6. Which AI providers are supported?

The README says the app supports more than ten providers and names options such as:

Different modules can be configured separately:

You do not necessarily need to use the same provider for every task.

For example:

7. Using it without programming knowledge

The repository provides desktop builds through GitHub Releases:

The user guide describes this general process:

The main external tools are FFmpeg, yt-dlp, and Deno.

8. Running from source

The documented requirements include:

Python 3.10 or newer
FFmpeg 4.4 or newer
a recent yt-dlp version
Deno 2.x

Basic developer setup:

git clone https://github.com/jipraks/yt-short-clipper.git
cd yt-short-clipper
pip install -r requirements.txt
python app.py

This route is intended for developers. Non-technical users should generally use a packaged desktop release.

9. How should YouTube cookies be treated?

The user guide explains that a cookie file can let the application access YouTube on behalf of a signed-in account.

A cookie file is not harmless. It may represent an authenticated browser session.

Security rules:

The repository documentation also warns that cookies expire and must never be shared.

10. Where are API keys stored?

The README says the application creates a config.json file on first run to store settings and API keys.

The current ConfigManager implementation writes configuration to a normal JSON file. Users should therefore treat the configuration file as sensitive data.

Avoid:

If a key is exposed, revoke it at the provider and create a replacement.

11. Strengths

A broad end-to-end workflow

Many repositories only identify timestamps or cut video. This project combines download, highlight selection, portrait reframing, hooks, captions, watermarking, and metadata.

Desktop interface

Packaged builds make the project accessible to people who do not want to run Python manually.

Multiple providers

Separate provider settings for different AI modules reduce dependence on one company.

Two speaker-tracking choices

Users can choose the speed of OpenCV or try MediaPipe for multi-speaker conversations.

Hardware acceleration

The project supports common hardware encoders from NVIDIA, AMD, Intel, and Apple.

12. Important limitations

Highlight selection is transcript-driven

Visual comedy, demonstrations, reactions, or action with little speech may not be selected correctly.

A highlight is not guaranteed to become viral

The model only judges the transcript according to its prompt. It does not know your audience with certainty.

Speaker tracking can fail

The largest face is not always the active speaker. Lip movement can also be affected by lighting, camera angle, masks, image quality, or people speaking off-camera.

Captions can contain errors

Names, technical terms, slang, accents, and noisy audio can cause transcription mistakes.

Hooks and metadata may exaggerate

AI-generated hooks and titles should be checked for misleading language, excessive clickbait, or claims not supported by the source.

API cost is variable

Longer videos, more clips, larger transcripts, and more expensive models increase cost.

External services can change

Changes to YouTube, yt-dlp, model APIs, or TTS services may break parts of the pipeline.

13. Copyright and platform rules

The repository disclaimer says the tool is for personal or educational use, users must respect YouTube's Terms of Service, and they must have rights to the content being processed.

Safer uses include:

Cropping a video, adding captions, or placing a watermark does not automatically give you rights to someone else's content.

14. Who is this repository for?

It can be useful for:

15. Who may not need it?

It may not fit your needs when:

16. A practical workflow

A sensible production workflow is:

This is safer and more reliable than fully automatic publishing.

17. Comparison with manual editing

The project is best treated as an editing assistant, not a complete replacement for a human video editor.

18. License

The repository uses the MIT License.

It allows broad use, copying, modification, distribution, and commercial use, provided that the copyright and license notice remain in copies or substantial portions of the software.

The software license does not grant rights to videos, music, images, trademarks, or other third-party content.

19. Conclusion

YT-Short-Clipper is a relatively complete desktop pipeline for converting long YouTube videos into vertical short-form clips.

The shortest explanation is:

Its main strength is reducing repetitive work. Its limitations are that AI does not fully understand every visual event, captions and crops can be wrong, APIs may cost money, and all outputs need human review.

SEO title suggestions

• What Is YT-Short-Clipper? An AI Tool for Turning YouTube Videos Into Shorts
• Beginner Guide to jipraks/yt-short-clipper for TikTok, Reels, and YouTube Shorts
• How YT-Short-Clipper Finds Highlights, Reframes Video, and Creates Captions
• Is This AI Podcast-to-Shorts GitHub Repository Worth Using?

SEO meta description

A beginner-friendly explanation of jipraks/yt-short-clipper: how AI selects highlights from transcripts, cuts video with FFmpeg, converts footage to 9:16 with OpenCV or MediaPipe, generates TTS hooks and Whisper captions, creates SEO metadata, and what users should know about cost, limitations, cookies, API keys, and copyright.

References

1. GitHub repository: https://github.com/jipraks/yt-short-clipper
2. README: https://github.com/jipraks/yt-short-clipper/blob/master/README.md
3. User Guide: https://github.com/jipraks/yt-short-clipper/blob/master/GUIDE.md
4. License: https://github.com/jipraks/yt-short-clipper/blob/master/LICENSE
5. Releases: https://github.com/jipraks/yt-short-clipper/releases