StoryShort — AI video creation for agents

StoryShort (https://storyshort.ai) turns prompts and scripts into finished short-form videos: script writing, AI visuals, voiceover, captions, rendering and publishing to connected social accounts. This skill drives the StoryShort MCP server over plain HTTP.

Setup

1. Create an account at https://storyshort.ai (Free plan works).
2. Generate an API key in Settings → API (https://storyshort.ai/app/settings/api).
3. Export it: STORYSHORT_API_KEY=ss_...

All calls go to one endpoint. Define this helper once per session:

ss() { # usage: ss <tool_name> '<json_arguments>'
  local args="${2:-null}"; [ "$args" = "null" ] && args="{}"
  curl -s -X POST https://api.storyshort.ai/mcp \
    -H "Authorization: Bearer $STORYSHORT_API_KEY" \
    -H "Content-Type: application/json" -H "Accept: application/json, text/event-stream" \
    -d "{\"jsonrpc\":\"2.0\",\"id\":1,\"method\":\"tools/call\",\"params\":{\"name\":\"$1\",\"arguments\":$args}}" \
    | sed -n 's/^data: //p' | jq -r '.result.content[0].text' | jq .
}

(Any MCP-capable client can instead add https://api.storyshort.ai/mcp directly as a streamable-HTTP server — OAuth or the same Bearer key.)

Discover before you spend

Everything is discoverable; never guess ids or prices:

ss get_account                 # plan + remaining credits — check this first
ss list_video_tools            # 20+ tools: faceless, ugc ads, movie maker, podcast, asmr, documentary...
ss get_video_tool '{"tool":"script_to_video"}'   # exact options for one tool
ss list_voices '{"query":"deep male"}'
ss list_caption_themes         # 33 styles: beast, neon, glass, comic, elegant...
ss list_image_styles
ss list_media_options          # media types + AI video tiers with credits per clip
ss list_music '{"query":"epic"}'
ss list_image_models           # standalone image models + costs
ss list_video_models           # Sora, Veo, Kling, Seedance 2.5 + per-second pricing
ss list_connected_accounts     # TikTok / YouTube / Instagram account ids for publishing

Create a video (always estimate first)

ss estimate_credits '{"tool":"script_to_video","script":"<the script>","mediaType":"images"}'
ss create_video '{"tool":"script_to_video","script":"<the script>","title":"My video","mediaType":"images","imageQuality":"basic","captionTheme":"neon","voice":"openai_ash","aspectRatio":"9:16"}'

Prompt-driven tools take "prompt" + "duration" (30/60/120/180 seconds) instead of "script". Other useful params: imageStyle (id or "auto"), videoTier (pvideo|pvideo-1080|seedance|seedance25 when mediaType is "videos"), language, dynamism, brandKitId, character, references, knowledgeSources.

Generation is asynchronous (1–5 minutes). Poll every 20–30 s:

ss get_video '{"videoId":"<id>"}'   # status: queued -> processing -> completed | failed

status: "completed" means the video is built and editable (share editorUrl with the user). For the final MP4:

ss render_video '{"videoId":"<id>"}'          # async — poll get_video until renderStatus is "completed" and url is set

Publish to social accounts

ss publish_video '{"videoId":"<id>","accountIds":["<from list_connected_accounts>"],"title":"Caption here","privacyStatus":"SELF_ONLY"}'
ss schedule_publish '{"videoId":"<id>","accounts":[{"id":"<id>","platform":"tiktok"}],"scheduledTime":"2026-09-01T18:00:00Z"}'

publish_video renders automatically if needed (can take minutes). Publishing posts to real accounts — confirm with the user before calling it, and prefer privacyStatus: "SELF_ONLY" on TikTok for tests.

Standalone images and clips

ss generate_image '{"prompt":"...","model":"nano-banana-2","aspectRatio":"9:16"}'
ss generate_clip '{"prompt":"...","model":"kling-v3","duration":5}'
ss get_media '{"mediaId":"<id>"}'   # poll until status "completed", then use .url

Image-to-image: pass "imageInput":["<url>"]. A generated image URL can seed a video via create_video's "startingImageUrl".

Workflow tips

• Call get_account first; refuse to start work that costs more credits than the balance and point the user to the upgrade URL instead.
• Always quote the estimate_credits price to the user before create_video on expensive configs (mediaType "videos", movie maker, podcast).
• One video at a time: create, poll to completion, then move on. Do not fire parallel create_video calls unless the user asked for variants (numberOfVideos handles that server-side).
• Titles you set are kept; leave title out to get an auto-generated one.
• Errors come back as readable text in the tool result (e.g. "Not enough credits (credits_missing)") — relay them, don't retry blindly.

Docs: https://storyshort.ai/mcp · REST API: https://storyshort.ai/api-docs · Support: https://storyshort.ai/contact