# Show HN: OpenMontage Drama Studio – self-hosted AI studio for short dramas

> Source: <https://github.com/charliewwdev/OpenMontage>
> Published: 2026-10-28 04:14:27+00:00

**From a one-line idea to a finished short drama, viral short or product clip.**

  Script → characters → scenes → props → storyboard & shots → voice → edit, in one self-hostable web studio.

⭐ If this is useful, a star helps others find it.

**English** · [简体中文](https://github.com/charliewwdev/OpenMontage/blob/main/README_zh-CN.md)

OpenMontage Drama Studio is a web studio for making short-form video with AI. You describe an idea in one
sentence (for example *"10-episode revenge drama, 9:16, hook in the first 3 seconds"*), the studio drafts the
script, and you move it through a production line where every stage is editable. Characters, scenes and props
become reference images that feed the video shots, so you can redo a single character or shot without starting over.

- **Bring your own keys.** Text models, image models and video models are configured per account in*Model settings* .
- **Browse first.** The inspiration wall, skills and asset pages can be browsed without signing in; signing in is
required for model calls and saving projects.
- **Step by step or one click.** Run each stage yourself, or let the agent run the whole pipeline in the background
and then re-run any stage or swap a version.

Try the hosted instance: [openmontage.dtok.io](https://openmontage.dtok.io/)

| # | Stage | What happens | 
|---|---|---|
| 1 | **Script** (剧本) | Genre, aspect ratio, visual style, synopsis or source material; generate, continue, rewrite, lock the episode | 
| 2 | **Characters** (角色) | Character art (Seedream, or ComfyUI on your own GPU), version history, turnarounds, reference re-draws, local upload | 
| 3 | **Scenes** (场景) | Space and lighting setup, scene concept art | 
| 4 | **Props** (道具) | Key props and material detail | 
| 5 | **Storyboard & shots** (分镜) | LLM shot breakdown; Seedance generates each shot using the character / scene / prop references | 
| 6 | **Voice** (配音) | Narration and dialogue TTS (edge-tts by default, OpenAI-compatible speech as fallback), mixed in on export | 
| 7 | **Edit** (剪辑) | Order the segments on a timeline and export the episode | 

In the current UI, stages 1–4 have their own tabs and stages 5–7 happen together per segment in the
**Workspace** (工作台) tab. The **infinite canvas** shows the same seven stages as a node graph, coloured by progress.

- **Short dramas.** Project templates for 爽剧 (hype drama), 短剧 (short drama) and 正剧 (straight drama), or upload your own script.
- **Viral short videos.** The*Viral gallery* (爆款内容广场) has beat-by-beat playbooks for Douyin, TikTok, Xiaohongshu,
Instagram Reels, YouTube Shorts and Bilibili (hooks, pacing, CTA) that you can send straight into a project.
These are structural templates based on common feed patterns, not scraped trending charts.
- **Ad and product videos.** A built-in e-commerce conversion template (pain point → product demo close-ups →
benefit captions → CTA), a soft product-placement Reels playbook, and marketing / e-commerce skills in the
skill market. They run through the same workshop; there is no separate ad editor.

| Module | What it does | 
|---|---|
| Viral gallery (爆款内容广场) | Platform playbooks, copyable prompts, a visual reference wall | 
| Creation hub (创作中心) | Text / visual / audio / video / workshop / canvas entry points | 
| Creation agent (创作 Agent) | One-line project setup: genre, episode count, aspect ratio, hook; generates the first scripts and checks which models are configured | 
| Workshop (影视工坊) | Project library and the 7-stage production line | 
| Assets (个人资产) | Media and prompt library, filtered by character / scene / prop / storyboard / final cut | 
| Skill market (Skill 专家) | Official and community skills (episode scripts, shot prompts, character looks, director workflows) you can install into the agent | 
| Infinite canvas (无限画布) | The project pipeline as a draggable node graph | 
| Model settings (模型配置) | Text and media providers, connection test, hot reload without restart | 
| Also | Works channel, challenges, resources, usage stats, task queue, help center, credits, referral program | 

## **Screenshots**

```
  React UI (studio/src)
       │
       ▼
  FastAPI (studio/server) ──▶ PostgreSQL  (projects · studio_kv · meta)
       │            └──────▶ Redis       (media job queue)
       └──▶ MinIO / S3     (inspire · media blobs)
```

| Path | Contents | 
|---|---|
| `studio/` | Drama Studio front end and API | 
| `studio/config/platform.yaml` | Navigation, workshop stages, tools, copy (single source of truth) | 
| `pipeline_defs/` ·`skills/` | Agent pipelines and director skills | 
| `lib/` ·`tools/` | Tool registry | 
| `docs/screenshots/` | Screenshots used in this README | 

|  | Included in `docker-compose.studio.yml` ? | 
|---|---|
| Studio app (UI + API) and media worker | ✅ built from this repo | 
| PostgreSQL 16 | ✅ bundled | 
| Redis 7 (media job queue) | ✅ bundled | 
| S3-compatible object storage (MinIO / S3) | ❌ **you provide** an endpoint, access key, secret and bucket (`OSS_*` /`MINIO_*` /`S3_*` ) | 
| Text LLM key | ❌ **you provide** : DeepSeek, OpenAI, Claude, Gemini, Qwen (通义), Doubao (豆包), Kimi, GLM, MiniMax, OpenRouter or another OpenAI-compatible endpoint | 
| Image + video generation | ❌ **you provide** : a Volcengine Ark key for**Seedream** (images) and**Seedance** (video). These are**paid third-party APIs** . Images can instead go to your own ComfyUI server. | 
| TTS | edge-tts by default (no key); optional OpenAI-compatible speech endpoint | 

Other things to know:

- `studio/.env.example` ships with defaults for the hosted instance (`openmontage.dtok.io` , its MinIO host and the
DTok account API in`API_BASE_URL` ). Replace them with your own values. By default sign-in is proxied to the
account API in`API_BASE_URL` ; see`STUDIO_AUTH_REQUIRED` and the demo-user settings in the same file.
- The hosted instance has a credits / top-up page. Self-hosted, you pay your model providers directly.

Docker (Compose reads `.env` from the repo root):

```
cp studio/.env.example .env
# fill in DATABASE_URL (or keep the bundled drama-db default), your S3/MinIO settings, and model API keys
docker compose -f docker-compose.studio.yml up --build
# open http://localhost:8787 · health check: GET /api/health (Postgres + object storage)
```

Local development:

```
cp studio/.env.example studio/.env
# fill in DATABASE_URL, MinIO (OSS_*), provider keys, then export them into your shell

cd studio/server && pip install -r requirements.txt
uvicorn app:app --host 0.0.0.0 --port 8787

cd studio && npm install && npm run dev   # http://127.0.0.1:5173, /api is proxied to 8787
```

Agent / CLI pipeline notes: [`AGENT_GUIDE.md`](https://github.com/charliewwdev/OpenMontage/blob/main/AGENT_GUIDE.md).

Python · FastAPI · React · PostgreSQL · Redis · MinIO / S3 · Seedream · Seedance · edge-tts · FFmpeg · Docker / Coolify

OpenMontage Drama Studio is an independent product by [@charliewwdev](https://github.com/charliewwdev).
It started from the AGPL-3.0 [calesthio/OpenMontage](https://github.com/calesthio/OpenMontage) codebase and has
since been rebuilt into a separate short-drama studio. It is **not affiliated with or endorsed by** the upstream
project. The repository keeps the upstream git history, so upstream authors appear in the contributor list.

Licensed under the **GNU AGPL-3.0**; see [`LICENSE`](https://github.com/charliewwdev/OpenMontage/blob/main/LICENSE).

This is a modified version of [calesthio/OpenMontage](https://github.com/calesthio/OpenMontage). Original code is
copyright its respective authors; modifications are by @charliewwdev (see the git history for what changed and when).
If you run a modified version as a network service, AGPL-3.0 §13 requires you to offer its source to your users.