# How to Automate YouTube Shorts from Script to Upload > Automating YouTube Shorts means letting software do the slow, repeated parts of each video: writing the script, recording the voice, adding captions and uploading. The ideas and the judgement stay with you. Done well, it lets one person post on a steady schedule. Done badly, it floods a channel with copies that YouTube will not pay for. - URL: https://www.agentskillpacks.com/products/shortfactory-ai - Product: ShortFactory AI - Price: $29 one-time (USD), instant download, lifetime v1 updates - Updated: 2026-10-21 - Seller: ToolGenX (https://www.agentskillpacks.com) ## Quick answer To automate YouTube Shorts, chain five steps in code: an AI model writes a short script, a voice tool reads it, a script pairs the audio with visuals, a caption tool adds timed text and the YouTube Data API uploads the file. ShortFactory AI is a Python pipeline that does this for a one-time $29, plus your own API costs. ## What ShortFactory AI is ShortFactory AI is a Python pipeline that makes YouTube Shorts: Claude writes the script, Gemini generates the images, ElevenLabs reads the voiceover, Whisper adds synced captions, and the YouTube Data API uploads each video as private. Budget about $0.25 to $0.30 per Short in API fees at October 2026 prices. ## Why does making YouTube Shorts by hand eat your evenings? Making YouTube Shorts by hand eats your evenings because one short video still needs five separate jobs: a script, a voice track, visuals, captions and an upload. Each job uses a different tool. The video is short, but the work around it is not. Here is what that costs a faceless channel: - **One video, many apps.** You write in one place, record in another, edit in a third and upload in a fourth. - **The export queue.** Renders take time, and a small fix means exporting again. - **Broken streaks.** When one Short takes most of an evening, a busy week means no posts at all. - **The math stops working.** If each video costs hours, a steady schedule costs more time than one person has. Volume alone does not fix this. By 2026-10-14 my own YouTube channel had 330 videos and 55 subscribers at one upload a day, and 11 of the last 12 videos had between 1 and 7 views. Those videos came from NotebookLM, not from this pipeline. The lesson carried over anyway: making more videos faster only helps if each one has something worth watching. Shorts can also run longer than people think. YouTube now allows [Shorts up to three minutes long](https://support.google.com/youtube/answer/15424877), though many faceless formats stay well under one minute. ## How to make a YouTube Short by hand, step by step Making a YouTube Short by hand is six steps, from topic to upload, and each one can be done with free or low-cost tools. Do a few by hand before you automate anything, so you know what a good Short in your niche looks like. 1. **Pick one clear idea.** One Short, one point. "Three facts about X" works. "Everything about X" does not. Write the point as one sentence before you start. 2. **Write a short script.** About 75 to 120 words fits 30 to 45 seconds of speech. Put the hook in the first line, because viewers decide fast whether to keep watching. 3. **Record the voice.** Use your own voice, or a text to speech tool, which turns written text into spoken audio. Read the script once out loud first and cut any line you trip over. 4. **Add visuals.** Use your own footage or licensed stock clips. Set the frame to vertical, 9:16, which is the tall phone shape Shorts use. 5. **Add captions.** Captions are the words on screen, timed to the voice. They keep the Short readable with the sound off. Check the spelling of every name. 6. **Upload with care.** Write a plain title, add the right audience setting and fill in the AI use question in YouTube Studio if any part looks realistic and was made with AI. ### The YouTube rules that matter for automated Shorts Two YouTube policies decide whether an automated channel works long term. Read both before you automate a single step. First, monetization. YouTube's [channel monetization policies](https://support.google.com/youtube/answer/1311392) were updated in July 2025 to rename "repetitious content" as "inauthentic content". They say AI content built from generic templates that gives the impression of mass production is not eligible to earn money, unless the creator adds original insight. A slideshow of stock clips with a robot voice reading facts is the exact pattern they describe. Second, disclosure. YouTube asks creators to [disclose realistic AI content](https://support.google.com/youtube/answer/14328491), such as making a real person seem to say something they did not, or showing a realistic scene that never happened. AI used for scripts, captions or ideas does not need a label. YouTube does not list a generic AI narrator as needing a label, but realistic AI images of real places or events do. When in doubt, tick the altered or synthetic content box. My rule from these two pages is simple: one channel, a pace you can review, and something in each video that a template could not have written. ## How YouTube treats an API upload as a Short There is no separate Shorts upload. A script sends a Short through the same videos.insert call as any other video. YouTube's [Shorts upload help](https://support.google.com/youtube/answer/12779649) says a Short is up to 3 minutes long, with a square or vertical frame. Get the shape right and YouTube does the rest. The API also has a daily cap. Google's [quota page](https://developers.google.com/youtube/v3/determine_quota_cost) gives a project 100 videos.insert calls a day by default. I checked it on October 21, 2026. Search results show three routes to the same upload: | Route | How it works | Trade-off | | ---------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------- | | No-code template tool | One video template filled from a spreadsheet, linked to YouTube through Zapier, as in [Creatomate's guide](https://creatomate.com/how-to/automate-youtube-shorts) | Easy start, but one shared template can drift toward the mass-produced look the monetization policy warns about | | Workflow tools such as n8n | Nodes chain an AI script, a render step and an upload | Flexible; you build and maintain the flow | | A Python pipeline (this kit) | Scripts run each job in order on your computer | Needs Python and API keys; you can read and change every step | ## The faster way: let ShortFactory AI run the pipeline The faster way is to let a Python pipeline run the five production jobs in order, with you choosing the topics and checking the results. A pipeline is a chain of small programs where each one hands its output to the next, like a kitchen line. ShortFactory AI is that chain for YouTube Shorts. The repeated steps run on their own, while the choices stay with the person who owns the channel. It is built for one disciplined channel, not a farm of accounts. ### How to use it, step by step 1. **Buy and download.** Gumroad sends the files right after payment. 2. **Set up Python.** Install Python 3.10 or newer and ffmpeg, then run `install.bat` on Windows or `install.sh` on Mac and Linux. Both create a virtual environment and install the packages. 3. **Add your API keys.** A first dry run starts a setup wizard that asks for an Anthropic key and a Gemini key, plus an optional ElevenLabs key. YouTube access is a one-time sign-in through `scripts/setup_youtube_oauth.py`. The [Gemini API pricing page](https://ai.google.dev/gemini-api/docs/pricing) and [ElevenLabs API pricing page](https://elevenlabs.io/pricing/api) show what each call costs. 4. **Give it topics.** Pass a headline with `--news`, or let it suggest trending topics from Reddit, RSS feeds and Google Trends. Drop any idea that needs a human face. If your best ideas already live in a blog post, the [AI Content Empire Builder](/products/ai-content-empire-builder) pulls Shorts scripts and hooks out of it, and the [Podcast to Empire Kit](/products/podcast-to-empire-kit) does the same for episodes. 5. **Run the pipeline.** It writes the script with Claude, makes the voiceover with ElevenLabs, generates the visuals with Gemini, adds synced captions with Whisper, then uploads through the [YouTube Data API](https://developers.google.com/youtube/v3/docs/videos/insert). Run one test video first. Google has been retiring Gemini image models, and if the image call fails the pipeline falls back to plain colour frames. 6. **Review before it goes public.** Every Short uploads as private, so you watch it on YouTube and make it public yourself. If your Google Cloud project is new and unverified, YouTube keeps API uploads private until the project passes Google's API audit, as the [videos.insert reference](https://developers.google.com/youtube/v3/docs/videos/insert) notes. Until then, download the video and upload it in YouTube Studio. The pipeline does not schedule posts. If you schedule in YouTube Studio, check the setting twice: on 2026-10-20 a wrong scheduling field made one of my videos public at once, for about a minute. ### What the output looks like Each run produces a vertical Short of about 60 to 90 seconds, with a voiceover, three AI images with slow zoom and pan, word-by-word captions and background music. It uploads to your channel as a private video with a generated thumbnail and prints the link. If the pipeline crashes halfway, running it again picks up where it stopped. I have not published a Short from this pipeline on my own channel yet, so I describe the output instead of linking one. Keep YouTube's limits in mind: when I bulk uploaded in October 2026, one channel stopped at about 72 uploads in a day, and Studio still showed OK for videos that were never created. ### How much it costs and how much time it saves API fees per video are small but not fixed. The README gives about $0.11: $0.02 for the Claude script, $0.04 for the images and $0.05 for the voice. At October 2026 list prices that looks low. ElevenLabs Multilingual v2 costs $0.08 per 1,000 characters, so a 60 to 90 second script runs about $0.07 to $0.11 for the voice, and four Gemini images at $0.039 each come to about $0.16. Budget roughly $0.25 to $0.30 per Short and check your own bill after the first ten. I have not timed a Short by hand or through the pipeline, so I will not give you a ratio. What the pipeline removes is the switching between apps. Your time goes into topics and review. The time you get back is worth spending on the part a pipeline cannot do: better ideas and a real point of view in each video. ## By hand vs a generic AI prompt vs ShortFactory AI The Short is the same in all three columns: a script, a voice, visuals, captions and an upload. What changes is how many of those steps you do by hand, what each video costs and how much you still have to check. | | By hand | Generic AI prompt | ShortFactory AI | | --- | --- | --- | --- | | Time per Short | I have not timed it, so no number here | Script only; voice, edit and upload are still by hand | No timed run published yet | | Cost per Short | Your hours, plus any paid editing tools | Your existing AI plan | Roughly $0.25 to $0.30 in API fees at October 2026 list prices | | Voice | Your own voice or a voice tool by hand | None | ElevenLabs voice through the API, or your computer voice as a fallback | | Captions | Typed or fixed in your editor | None | Word-by-word captions from Whisper, run on your computer | | Upload | By hand in YouTube Studio | By hand | Through the YouTube Data API, as a private video | | Setup | None | None | Python 3.10 or newer, ffmpeg and API keys; follow the guide once | ## What is inside ShortFactory AI ShortFactory AI is a Python pipeline that takes a topic and returns a private YouTube Short, with Claude for the script, Gemini for the images, ElevenLabs for the voice and Whisper for captions. - A full Python pipeline: research, script, visuals, voiceover, captions, music and upload - Claude for the script and ElevenLabs for the voiceover, in English or Hindi - Three Gemini images per video with slow zoom and pan, plus a Gemini thumbnail - Word-by-word captions from Whisper, burned into the video - Upload to YouTube through the YouTube Data API, private by default - Topic discovery from Reddit, RSS feeds and Google Trends - A setup prompt you can paste into any AI assistant for step by step help ## Who ShortFactory AI is for, and who should skip it This pipeline fits one person running one channel who is limited by production hours, not ideas. It does not fit anyone planning to mass post across many accounts, which is what gets channels removed. - For you if: You know which niche you would run if each Short took minutes, not hours. - For you if: You have broken a posting streak because editing ate the week. - For you if: You run a faceless channel where production time, not ideas, is the limit. - For you if: You can follow a Python setup guide and paste in a few API keys. - Skip it if: You run a channel built on your own face and stories. That needs you on camera, not a pipeline. - Skip it if: You plan to mass post from many accounts. That breaks YouTube rules, and this is built for one channel. - Skip it if: You will not touch Python or API keys. The pipeline needs both to run. ## FAQ ### Can you fully automate a YouTube Shorts channel? You can automate the production steps, but you should not automate the judgement. Software can write a draft script, voice it, caption it and upload it. Picking topics, checking facts and adding your own angle still need a person, and YouTube's monetization rules reward exactly that human part. ### Do AI-made Shorts get monetized? They can, if they are not mass-produced or repetitive. YouTube's inauthentic content policy excludes AI content made from generic templates that look mass-produced, unless the creator adds original insight. Each Short needs something real to say, not just a new topic in the same mold. ### Is $0.11 per video the real cost? Probably not today. The README gives about $0.11, but at October 2026 list prices the voice alone runs about $0.07 to $0.11 and four Gemini images about $0.16. Budget roughly $0.25 to $0.30 per Short, and check your own bill after the first ten videos. ### Do I need coding skills? You need to be able to follow a Python setup guide: install Python and ffmpeg, run an install file and paste API keys when the setup wizard asks. You do not need to write code. If a terminal makes you nervous, this is the wrong tool. ### Does it work with ChatGPT? Not as the engine. The pipeline runs as Python on your own machine and calls Claude for the script, through an Anthropic API key or the claude command line tool on a Max plan. It also needs a Gemini key and YouTube access, and ElevenLabs is optional. ChatGPT can still help with setup: the included setup prompt works in any AI chat. ### Is it a one-time payment? Yes, ShortFactory AI costs $29 once. There is no subscription. You pay the API providers directly for what you use, and the README has a cost table per service. ### Will my channel get banned for automated uploads? Uploading through the YouTube Data API is an official, allowed route. What gets channels in trouble is spam: many accounts, copied videos or misleading content. Run one channel, post at a pace you can review and follow YouTube's policies, and you avoid the patterns that trigger action. If your Google Cloud project is new and unverified, YouTube keeps API uploads private until the project passes Google's API audit. Until then, download the video and upload it in YouTube Studio. ### Do I have to label AI content on YouTube? You must disclose AI content that looks realistic, such as a real person saying something they did not say or a realistic scene that did not happen. AI used for scripts or captions does not need a label, and YouTube does not list a generic AI narrator as needing one. This pipeline makes AI images, so a realistic image of a real place or event does. When in doubt, tick the altered or synthetic content box.