BokeBox

Content in. Private podcasts out.
_{Turn multi-source content into private shows you can finish. Persona, voice, plugins, MCP — self-hosted.}

English · 简体中文

Demo · Docs · Product Hunt · Get started · UI · Features · Disclaimer

---

Bottom line first

BokeBox turns “I’ll read/watch it later” into private episodes you can actually finish — and keep.

You put in	You get back
Video / links / drafts / meetings & courses	Spoken episode + cover + flashcards + progress
Global or per-episode persona / voice	Feels like a show made for you
Plugins / MCP / schedules	Extensible inputs, AI-native control

One line: a self-hosted private AI podcast studio — your data stays local, the pipeline is pluggable, listening quality beats feature bloat.

Try it now

• Demo: https://bokebox.aiuo.net
• Docs: https://bkb-docs.aiuo.net
• Local in three steps:

git clone https://github.com/vastsa/BokeBox.git
cd bokebox
cp .env.example .env   # add API keys
./start.sh             # http://localhost:5173

Docker (recommended prebuilt image):

cp .env.example .env
docker pull ghcr.io/vastsa/bokebox:latest
./start.sh docker      # http://localhost:8787

---

30-second pitch

  What you put in                    What BokeBox gives back
  ─────────────────                  ──────────────────────
  Meetings / notes                   Rhythmic spoken episodes
  Long reads / talks / courses       Host persona + natural voice
  Any link (plugin-extensible)       Cover · flashcards · progress

1. Drop it in — video, link, draft, or plugin sources
2. Tune optionally — persona, voice, prompts
3. Let it run — ASR → spoken script → TTS → cover / cards
4. Listen — player + albums + star map + review

---

Why it matters

Pain	What BokeBox does
Endless “read later”	Moves content into ear time
Robot read-aloud	Rewrites into spoken structure with persona
Heard and forgotten	Auto flashcards for key points
Locked-in inputs	Source / ASR / TTS / Schedule plugins
Cloud lock-in	Single-user self-host; jobs & media stay local

Good fit: knowledge workers, commute listeners, private “shows for myself”, privacy-minded self-hosters.
Not a fit: public podcast platforms, multi-tenant SaaS, live collaborative editing.

---

UI

Create, listen, and manage assets in one private space.

Home	Player
Flashcards	Star Map

Settings

Persona	Prompts
Plugins	MCP
Schedules	Site

---

Feature map

Expand full checklist (pipeline / listening / plugins / MCP / deploy)

Multi-source input

• Local upload: video / audio / draft
• URL import: article body, public media links
• Pin a Source plugin or auto-match; assign album, persona, voice on create

AI production pipeline

• Extract → ASR → spoken script → cover / notes / flashcards → TTS
• Async jobs with home progress; re-run from a chosen step and skip finished stages
• Publish to library, retry failures, delete jobs

Persona · voice · prompts

• Global host persona + per-episode override
• Preset voices + text-described voice design
• Prompt hub: cover / spoken / rewrite / flashcards with {{placeholders}}
• Content language: global default or per job

Assets & listening

• Job detail: transcript, script, notes, flashcards, cover, audio
• Player: progress memory, speed, sleep timer (incl. end of episode)
• Albums for continuous play; Star Map by tags

Settings center

• Voice / Persona / Prompts / AI service / Plugins / Schedules / MCP / Site / Account
• Account: UI language (system / zh / en) + theme (system / light / dark)

Schedules

• Schedule plugins discover candidates; Source plugins fetch/parse
• Built-ins: RSS, URL list, GitHub Trending, Hacker News, …
• Dedup, rate limits, run now / force, run history
• Guide: docs/guide/schedule.md

Plugins

• Shared Source / ASR / TTS / Schedule contracts; rescan / zip upload / enable in Settings
• Docs & examples: docs/plugins/ · docs/development/ · examples/

MCP (AI-native control)

• Built-in POST /mcp with auto long-lived token
• Copy Cursor / Claude / Codex install configs from Settings
• Tools include create_podcast_from_url / create_podcast_from_text / list_jobs / get_job / schedule tools, …
• Optional PUBLIC_BASE_URL for reverse-proxy install URLs

Cursor example:

{
  "mcpServers": {
    "bokebox": {
      "url": "http://localhost:8787/mcp",
      "headers": {
        "Authorization": "Bearer <token from Settings>"
      }
    }
  }
}

Deploy & privacy

• ./start.sh dev · ./start.sh prod single port · Docker prebuilt / local / China mirrors
• Single-user self-host: SQLite + local storage
• License: LGPL-3.0 · repo: https://github.com/vastsa/BokeBox

---

Get started

Prefer clicking first? Demo · details in docs

git clone https://github.com/vastsa/BokeBox.git
cd bokebox
cp .env.example .env   # add your API keys
./start.sh             # http://localhost:5173

First launch walks through account setup and model config.

Mode	Command	Open
Local dev	./start.sh	http://localhost:5173
Single-port prod	./start.sh prod	see script output
Docker prebuilt	docker pull ghcr.io/vastsa/bokebox:latest && ./start.sh docker	http://localhost:8787
Docker local build	./start.sh docker.local	http://localhost:8787
China mirrors build	./start.sh docker.cn	http://localhost:8787

---

Docs & links

Entry	URL
Online docs (EN/ZH)	https://bkb-docs.aiuo.net
Demo	https://bokebox.aiuo.net
In-repo docs	docs/ (VitePress)
Plugins / developer guides	docs/plugins/ · docs/development/
CI / images	docs/ops/ci-cd.md

pnpm docs:dev
pnpm docs:build
pnpm docs:preview

---

Roadmap & contribute

• Better multi-episode / continue-listening flows
• Richer voices and providers
• Subscription export (e.g. RSS) into apps you already use
• Lighter one-click desktop packaging

If that resonates:

1. ⭐ Star the repo
2. Open an Issue — what should be “podcast-ified”?
3. PRs welcome — UX, copy, voices & model adapters especially

Shareable one-liner: BokeBox turns videos, links, articles, meetings, and courses into private podcasts. Customize persona and voice, with MCP and pluggable sources. Self-hosted — your data stays under your control.

Content in. Private podcasts out. Most tools help you produce content faster; BokeBox helps you digest it better.

---

Appendix: requirements · commands · config · stack

Requirements

• Node.js ≥ 22.5 · pnpm 9.x
• OpenAI-compatible API (Chat / ASR / TTS; image model optional)

Common commands

Command	Description
./start.sh	Local dev (web 5173 + API 8787)
./start.sh prod	Build and run on a single port
./start.sh docker	Pull ghcr.io/vastsa/bokebox:latest and start
./start.sh docker.local	Build from local Dockerfile and start
./start.sh docker.cn	Build with China mirrors and start
./start.sh docker:down	Stop containers

Config highlights (.env)

OPENAI_API_KEY=sk-your-key
OPENAI_BASE_URL=https://api.example.com/v1
OPENAI_CHAT_MODEL=mimo-v2.5
OPENAI_TRANSCRIBE_MODEL=mimo-v2.5-asr
OPENAI_TTS_MODEL=mimo-v2.5-tts
OPENAI_TTS_DEFAULT_VOICE=冰糖

Full variables: .env.example.

Pipeline (sketch)

multi-source input (video / link / draft / plugins)
  → normalize → ASR/understanding → spoken script
  → parallel: cover / flashcards / TTS → library

Stack

React · Vite · Fastify · SQLite · ffmpeg · pnpm monorepo

License

LGPL-3.0 — open source; derivative works of the library must remain LGPL-compatible. Repo: https://github.com/vastsa/BokeBox

---

---

Disclaimer

This project is provided “AS IS”, for learning, research, and technical exchange only. The authors and contributors make no warranties, express or implied, including but not limited to merchantability, fitness for a particular purpose, non-infringement, accuracy, availability, or continued availability.

By using, deploying, modifying, or distributing this project (including derivatives and third-party plugins), you acknowledge and agree that:

1. Compliance is your responsibility
   You must comply with all applicable laws and regulations in your jurisdiction(s), including copyright and related rights, personal data / privacy, cybersecurity, generative-AI rules, trade secrets, platform terms of service, and export controls.

2. Content rights are on you
   You alone must ensure that materials you ingest, scrape, transcribe, rewrite, synthesize, store, or distribute are lawfully authorized or otherwise permitted. Do not use this project for infringement, circumvention of technical protection measures, unauthorized access, fraud, harassment, or other unlawful purposes.

3. Risks stay with the operator
   To the maximum extent permitted by law, the authors and contributors are not liable for any direct, indirect, incidental, special, punitive, or consequential damages (including data loss, downtime, account bans, third-party claims, or administrative penalties) arising from use of this project. Enabling plugins, MCP, schedules, guest access, reverse proxies, or public exposure is at your own risk.

4. Third-party services & dependencies
   Model APIs, TTS/ASR providers, fetch-style plugins, container images, and other third-party components have their own terms and billing. Their availability, cost, safety policies, and compliance duties are outside this project’s responsibility.

5. Demo & docs
   The demo instance and documentation examples are for illustration only and are not a product commitment, professional advice, or implied license. Do not put real secrets, private data, or protected content into the demo environment.

6. Open-source license
   Software licensing is governed by the root LICENSE (LGPL-3.0-only). This disclaimer supplements risk notice and liability boundaries and does not alter rights or obligations under LGPL. If any part conflicts with mandatory law, that part yields to the law and the remainder continues in effect.

If you do not agree, do not download, install, use, or distribute this project.

BokeBox
_{Private AI podcast box · Open Source · LGPL-3.0
github.com/vastsa/BokeBox}