Start from a story package
A short script with scene timing and one visual prompt per scene. That single file drives the whole pipeline.
ExperimentalAI short-form video pipeline for Windows
Turn a written story into a finished vertical short
FastShorts takes a script with scene prompts and produces a narrated, captioned 9:16 video with generated visuals, motion, music, and SFX.
FastShorts is a local Windows pipeline for faceless short-form video. Scene images come from AI image models, motion comes from local image-to-video generation, narration comes from cloud or local TTS, and the final MP4 is composed on your machine with word-level karaoke captions and ducked background music.
the-mirror-was-early/story.json6 scenes
ImagesMotionNarrationCaptionsMusic
Renderedthe-mirror-was-early.mp4 · 9:16 · 45s
How it works
From story package to finished MP4
One script drives every stage. Each stage runs locally, checkpoints its work, and hands off to the next.
Generate scene visuals
One vertical image per scene, generated in the cloud or fully locally through a ComfyUI text-to-image workflow.
Narrate and time the words
Narration comes from ElevenLabs or a local Kokoro TTS server, with word-level timing extracted for captions.
Animate each scene
Still images become motion clips through a local ComfyUI Wan image-to-video workflow, scene by scene, with resumable checkpoints.
Compose the final short
Clips are cut against the narration, word-level karaoke captions are burned in, and music is mixed under the voice with ducking. Out comes a finished 9:16 MP4.
Features
A whole shorts studio, running on your PC
FastShorts is not an editor. It is a production pipeline: writing stays yours, and everything after the script is automated.
Story packages in, videos out
The input is a script with scenes, timing, and prompts. The output is a finished vertical MP4. Everything between is pipeline.
AI images per scene
Scene visuals from cloud image models or a local ComfyUI text-to-image workflow when you want zero cloud dependency.
Local image-to-video motion
Scenes are animated on your own GPU with a ComfyUI Wan image-to-video workflow instead of a metered cloud renderer.
Cloud or local narration
ElevenLabs voices when you want them, or a fully local Kokoro TTS server in Docker when you do not.
Word-level karaoke captions
Captions are timed to each spoken word and burned into the video, in the style short-form viewers expect.
Music and SFX with ducking
Background music and effects are mixed under the narration automatically, including optional local audio generation.
Resumable renders
Completed scenes are checkpointed and skipped on a rerun, so a failed scene does not restart the whole video.
Built to be agent-driven
A companion app exposes the pipeline through a visible UI and a headless local HTTP API that an AI agent can drive end to end.
Local-first
Your renders happen on your hardware
No render queue. No per-video credits. No subscription meter.
Cloud shorts generators charge per render and keep your pipeline on their servers. FastShorts runs image-to-video, caption burning, and audio mixing on your own GPU through ComfyUI and FFmpeg. Cloud services are optional ingredients, like ElevenLabs voices, not the factory.
A pipeline, not an editor
There is no timeline to drag clips around. FastShorts turns a finished script into a finished video. If you want to cut existing footage into clips, that is FastClip.
Experimental
FastShorts is the most experimental tool in the Fast Series. The source is currently private and the pipeline is still taking shape, so there is no public download yet.
Compare
Where FastShorts fits
Cloud generators sell renders by the credit. Editors want your time. FastShorts wants a script and a GPU.
FastShorts is a good fit if you want
- Story-driven faceless shorts from a written script
- Renders on your own GPU instead of per-video credits
- Word-level karaoke captions burned in automatically
- Narration from ElevenLabs or a fully local TTS
- Music and SFX mixed under the voice without an editor
- A pipeline an AI agent can drive end to end
Other tools are still better if you need
- Clips cut from footage you already recorded (FastClip)
- A hands-on timeline editor with manual control
- A finished product today: FastShorts is experimental
- Mac or Linux support
- A hosted, no-setup cloud service
- Photorealistic long-form video generation
Facts
FastShorts facts
A quick reference for the essentials.
| Product | FastShorts |
|---|---|
| Developer | Calvin Sturm (Sturm Technologies LLC) |
| Category | AI short-form video pipeline |
| Platform | Windows 10 and later, 64-bit |
| Status | Experimental, no public download yet |
| Input | Story packages: a script with scene timing and one visual prompt per scene |
| Output | Vertical 9:16 MP4 with narration, word-level karaoke captions, music, and SFX |
| Visuals | AI images per scene, animated locally with a ComfyUI Wan image-to-video workflow |
| Narration | ElevenLabs voices or a local Kokoro TTS server |
| Processing | Local through FFmpeg and ComfyUI on your own GPU; cloud steps are optional |
| Technology | Windows companion app with a local HTTP API, FFmpeg, ComfyUI |
| Best for | Faceless story channels producing narrated vertical shorts from scripts |
| Not designed for | Editing existing footage (FastClip) or hands-on timeline editing |
FAQ
FastShorts questions
What is FastShorts?
FastShorts is a local Windows pipeline that turns a story package, a script with scene timing and visual prompts, into a finished short-form vertical video with AI-generated visuals, narration, word-level karaoke captions, music, and SFX.
Does FastShorts run locally?
The pipeline itself runs on your machine: composition through FFmpeg, image-to-video through ComfyUI on your own GPU, and optional local narration, image, and audio generation. Cloud services like ElevenLabs or cloud image models are optional steps, not requirements.
Is FastShorts for faceless channels?
That is the core use case: story-driven, narrated vertical videos produced without filming anything. If you have scripts and a Windows PC with a GPU, FastShorts is designed to turn them into finished shorts.
How is FastShorts different from FastClip?
FastClip starts from footage you already have and cuts it into clips. FastShorts starts from a written story and generates the video: images, motion, narration, captions, and music. They cover opposite ends of the short-form workflow.
What does FastShorts need to run?
Windows with PowerShell, FFmpeg, and a local ComfyUI install for image-to-video generation. Optional pieces include Docker for local Kokoro TTS narration and a local audio-generation setup for music and SFX.
When can I use FastShorts?
FastShorts is experimental and the source is currently private. There is no public download yet. FastCast, FastPlay, FastClip, and FastCompress are available today, and FastShorts will be announced on the Fast Series page when it is ready.
FastShorts is taking shape
While FastShorts is experimental, the rest of the Fast Series is downloadable today: FastCast, FastPlay, FastClip, and FastCompress are all available now as native Windows tools.