AI short drama & video creation
Turn a script into characters, a storyboard, keyframes and video clips with one workflow API, or pick image, video, speech and music models yourself. Every step is billed at the listed model price.
Short dramas and story videos need several kinds of model working together: a text model to break the script down, an image model for characters and keyframes, a video model to animate each shot, and voice or music on top. BigModel.io serves all of them with one key and one USD balance, and adds a workflow API that chains them for you.
Five steps, one request
- 01Find the charactersCharacters and their look, from the script.
- 02Write the storyboardScenes cut into 4–15 second shots.
- 03Character sheetsOne reference image per character keeps them consistent.
- 04First framesAn image model draws each shot's first frame.
- 05Video clipsA video model animates each frame, with native audio.
- 2
- characters
- 2
- clips
- ~6 min
- script to video
- $0.42
- total (quoted $0.60)
What you can build
The multimodal lineup
From script to clips in one call
POST /v1/workflows/short-drama takes a script and returns a workflow ID right away. The workflow extracts the characters and draws a reference image for each, writes a storyboard, renders a first frame per segment, then animates every frame into a clip. Poll GET /v1/workflows/{id} to read the characters with their images, the storyboard, and each segment with its frame and video URLs as they finish.
curl https://bigmodel.io/v1/workflows/short-drama \
-H "Authorization: Bearer $BIGMODEL_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: episode-1" \
-d '{
"script": "A courier finds a letter addressed to herself, dated ten years from now...",
"style": "cinematic, warm evening light",
"aspect_ratio": "9:16",
"max_segments": 6
}'
# Poll until status is succeeded; cancel with POST …/cancel
curl https://bigmodel.io/v1/workflows/WORKFLOW_ID \
-H "Authorization: Bearer $BIGMODEL_API_KEY"| Field | What it does |
|---|---|
script | Required. The story, up to 20,000 characters. |
style | Optional visual style applied to every image and clip. |
aspect_ratio | 9:16, 16:9 or 1:1. |
image_model | Text-to-image model for characters and keyframes. Default qwen-image-2.1. |
video_model | Image-to-video model for the clips. Default seedance-2.0-fast. |
max_segments | Number of storyboard segments. Default 6, at most 20. |
stop_after | storyboard, frames or videos (default): stop early to review before paying for clips. |
audio | Ask the video model for sound, on models that generate audio. |
dry_run | Return a price estimate without running anything. |
- Every step is billed at the listed price of the model it uses; failed steps are not charged. Use dry_run for an estimate first.
- Send an Idempotency-Key header: retrying with the same key returns the same workflow instead of starting a new one.
- Cancel a running workflow with POST /v1/workflows/{id}/cancel. Steps already finished stay billed.
Short-drama workflow guideRequest and response fields, stages, polling and errors.
Read the guideCreate an API keyRecommended model pairs
Prefer no code?
RooYunAn AI short-drama platform where an agent works on an infinite canvas for you: from the script to the storyboard, characters, and batches of images and videos, through to a finished episode.
Try RooYunAbout RooYun