Shorts Factory Pro: CSV to 30 TikToks & Shorts

Thirty videos is a CSV with thirty rows

Most faceless-video tools charge per video or per minute, and most free ones stop at a script. This kit goes the whole way on your own machine: your agent writes a short script for every topic in a fixed, checkable format, one command renders all of them into finished 1080x1920 videos, and a second command writes the captions, titles and posting schedule.

What it looks like

The first seconds of three sample videos the kit rendered from three rows of a CSV: a numbered list, a before/after, and a countdown. The word being spoken lights up in the captions.

What comes out

For every topic

• final.mp4 - 1080x1920, 30 fps, voice normalised to -16 LUFS
• Captions where the word being spoken lights up, placed clear of every platform's buttons
• One of five on-screen layouts per cut: fact, numbered list, countdown, quote, before/after
• A slow push-in so still frames never sit dead, and a thin progress bar across the top
• cover.png, captions.srt, and qa.json with every warning

For the batch

• captions.json - TikTok caption, Instagram caption, YouTube title and description, hashtags
• schedule.csv - one video a day, or every N days, from the date you choose
• qa_report.csv and UPLOAD_CHECKLIST.md

Measured, not promised

On the machine it was built on, three test videos of 18-24 seconds (two in English, one in Japanese) rendered in 35 seconds in total. Speed depends on your computer.

The QA report catches what loses viewers

• A hook that takes more than 3.5 seconds to say, or opens with a greeting
• A total outside 15-35 seconds, or a single cut longer than 7 seconds
• Speech too fast for the captions to be read
• On-screen text that had to shrink to fit

It also refuses to render any script that still contains [CHECK], so a line with an unverified number cannot slip into a finished video.

Voices

• edge-tts (default): hundreds of voices across dozens of languages, with captions that follow the real word timings. It sends the text of each line to Microsoft's online voice service to get the audio back, so it needs an internet connection.
• VOICEVOX: a local Japanese engine; nothing leaves your computer
• Your own recordings: one file per cut; nothing leaves your computer

What you need

Python 3.8+, pip install pillow edge-tts, and ffmpeg. A doctor command checks everything. No Node, no editing app, no API key, no account. Windows, macOS and Linux. Rendering happens on your computer; the default voice engine is the one part that goes online.

Honest limits

• It renders; it does not post. Nothing is uploaded or scheduled on any platform for you.
• No promises about views, followers or income. The QA checks catch known problems; they do not predict performance.
• edge-tts is an unofficial client for Microsoft's online voices: it needs internet, its availability can change, and you should check Microsoft's terms for commercial use. Each VOICEVOX voice has its own terms. Your own recordings are always yours.
• Use only images and music you have the rights to. The kit does not download stock footage or anyone else's content.

From the same publisher

• Shorts Script Timer (Free) - check a script's spoken length before you record: https://capafy.ai/agent/3315181728
• Retention Check (Free) - is a published short worth rebuilding? https://capafy.ai/agent/3675277146