ShortsAIHow to make AI videos, and which of the two methods you need
"AI video" means two different things: AI editing of footage you have (clipping, cropping, captions, dubbing), and AI generation of footage you do not have (script, actor, voice, b-roll).
Editing is cheap and reliable, and free with ShortsAI. Generation costs roughly $0.65 to $2.00 per short video through third-party models and is less predictable.
For most creators the highest-yield first step is clipping existing long videos; generation is for marketing content with no source footage.
What does "make an AI video" actually mean?
Two workflows share the phrase. In the first, you have a recording and AI does the editing: finding the moments, cropping to vertical, writing captions, dubbing the audio. In the second, you have no recording and AI produces one from a text description: it writes a script, synthesises a presenter and a voice, and assembles b-roll. The first is a mature technology with predictable output. The second is impressive, priced per video, and still needs a human to reject the misses. Decide which you need before choosing a tool.
Method 1: make AI videos from footage you already have
- Start with something long and conversational: a podcast, an interview, a webinar, a stream. Conversation clips well; screencasts do not.
- Upload it or paste the YouTube link into ShortsAI.
- The pipeline transcribes with faster-whisper, detects scenes, and has Google Gemini score the transcript for the strongest 3 to 15 moments.
- Each moment is reframed to 9:16 with MediaPipe face tracking, given a hook overlay and burned-in subtitles.
- Optionally dub any clip into 30+ languages with ElevenLabs, with subtitles regenerated to match.
- Download or publish to TikTok, Reels and Shorts.
Cost: free, with your own Gemini key (free tier 1,500 requests a day). Time: 5 to 8 minutes per 8-minute source on a CPU, about 50 seconds on a GPU.
Method 2: generate AI videos from a description
- Describe the product, service or business, or paste its URL.
- The AI writes a 30 to 60 second script with a hook and a call to action. Edit it.
- Choose an AI actor or upload a photo; the actor is generated with lip movement matched to an ElevenLabs voiceover.
- B-roll, subtitles and a hook overlay are added automatically.
- Preview, regenerate any part that missed, and export.
Cost: roughly $0.65 per video in Low Cost mode and $2.00 in Premium, billed by fal.ai and ElevenLabs. Expect to regenerate some outputs; budget two to three attempts per finished video when planning a batch.
Where do AI videos go wrong?
- Clipping a source with no speech. The moment detection is transcript-based; music and silent footage give it nothing to score.
- Trusting the crop on group shots. Face tracking follows one subject. For panels, use the split layout or GENERAL mode.
- Generated hands and props. Talking-head generation is good; an actor holding your product is not. Show the product in b-roll instead.
- Unlabelled AI content. Every major platform requires disclosure of synthetic video. Label it.
Do you need a GPU to make AI videos?
Not for clipping; the CPU path works on any machine with Docker and 8GB of RAM, just slower. Not for generation either, because the heavy models run at the provider. A GPU is not something you need to think about.
Common questions
- Can I make AI videos for free?
- Clipping and captioning existing footage is completely free with ShortsAI. Generating new video with AI actors uses paid third-party models at roughly $0.65 to $2.00 per video.
- What is the easiest way to make an AI video?
- Paste a YouTube link of a video you own into a clipper and let it return subtitled vertical clips. It needs no script, no footage planning and no prompt engineering.
- Can AI make a video from text only?
- Yes. The AI Shorts tool writes a script from a description and generates a presenter, voice, b-roll and captions from it.