shorts-studio
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| FFMPEG_PATH | No | Path to a custom ffmpeg binary | |
| PEXELS_API_KEY | No | Free API key from pexels.com/api, provides the best footage. | |
| SHORTS_BGM_DIR | No | A folder of royalty-free music; a random track is mixed in at 15% volume | |
| PIXABAY_API_KEY | No | Free API key from pixabay.com/api/docs | |
| SHORTS_FONTS_DIR | No | Directory with extra fonts for subtitles | |
| SHORTS_OUTPUT_DIR | No | Output directory for generated videos (default: ~/Videos/shorts-studio) |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Shorts Studio turns a topic into a narrated, subtitled short video (TikTok / Reels / YouTube Shorts). You (the model) write the script and the stock-footage search terms; the tools do voice, footage, subtitles and rendering. Script rules: start with a strong hook sentence, plain spoken text only (no markdown, emojis, headings, scene labels or "welcome to"), ~130-160 words for a 60-second short, in the user's language. Search terms: 3-8 short, concrete, visual English phrases (e.g. "city traffic night", "doctor laptop"), in the order the script mentions them. Then call make_short_video. Rendering takes 1-5 minutes. Report the video path, then suggest a title, description and hashtags, and mention credits.txt.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| make_short_videoA | Render a narrated, subtitled short video from a script you wrote: free neural voice-over (Microsoft Edge TTS), stock footage for each search term (Pexels/Pixabay with free keys, else Wikimedia Commons), word-highlight subtitles, optional background music, MP4 output. Takes 1-5 minutes. |
| generate_voiceoverC | Free neural text-to-speech (Microsoft Edge voices, 300+ voices, 70+ languages) to MP3, plus an SRT subtitle file timed to the words. |
| list_voicesA | List the free Edge neural voices, optionally filtered by language prefix (e.g. "ar", "en-GB") and gender. |
| search_footageA | Preview stock clips for a search term (no download): source, size, duration, author and licence. |
| doctorA | Check ffmpeg, footage providers, output and cache folders. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 5 tools
Each tool has a largely distinct role: doctor is health-check, list_voices is discovery, search_footage is preview, generate_voiceover is an audio artifact, and make_short_video is the full pipeline. The only soft overlap is that make_short_video internally performs voiceover and footage steps, so an agent must infer that the single-purpose tools are building blocks or previews rather than alternatives — the descriptions clarify this, but only after reading.
Four of five names follow a clean verb_noun pattern (make_short_video, generate_voiceover, list_voices, search_footage). 'doctor' is a bare noun that breaks the convention, though it is a recognizable idiom for a diagnostics command.
Five tools is well-scoped for a video-generation server: one diagnostics, one orchestrator, and three focused helpers. Nothing feels padded or redundant.
The core lifecycle — check environment, pick voice, generate audio, scout footage, render video — is covered with no true dead ends. Minor gaps: search_footage only previews with no standalone download tool, and there is no way to list or supply background music, validate a script, or generate audio from an existing file.