Babel Tower Logo

Babel Tower

Choose your role · enter the matrix

AI-powered creation, multilingual reach | Babel Tower — give every word voice, warmth, and a global echo.

About Babel Tower

Babel Tower is a next-generation AI-powered creation and reading platform for creators and readers worldwide. It is built around lightweight convenience, intelligent efficiency, and global reach—no complicated setup; sign in quickly with email and password.

We integrate AI-assisted creation, automatic multilingual translation, TTS audiobook generation, and unified audio/video tools so creators can ship quality work without burning out, generate multilingual editions and audiobooks in one click, and reach a global audience—while readers break language barriers, enjoy text or audio, and get a richer, more flexible reading experience.

Unlike traditional web-fiction platforms, we do not bind your copyright or impose harsh gatekeeping. Creators keep full control of their work; every creation can reach a global audience—reading and creation, without borders.

Development progress

Newest first

August 6, 2026 (day 132)

Progress

  1. Aura quality-eval public toggle: after evaluation, authors can publish results to the library quality board; readers see full per-dimension scores (progress bars + scores); unchecking removes the work from the board immediately.
  2. Library quality board fix: the rankings UI now correctly reads and renders dimension scores for publicly shared evaluations.
  3. Aura eval compute offload: the Hong Kong site forwards evaluation requests to the stronger local Aura service and writes results back.
  4. Aura eval upstream switch: fixed 403 from insufficient OpenAI-compatible proxy quota; switched to a faster available model to avoid reasoning-model timeouts and temperature-parameter limitations.

August 5, 2026 (day 131)

Progress

  1. Multilingual translation follows the reader-visible novel version (no false “chapter not in this work” after switching books) and uses the author’s BYOK key plus the selected / account-preferred model.
  2. Translation quality gates: reject chain-of-thought drafts and English mixed with Chinese; refuse to save bad output. Fixed unclosed heading HTML that made reader font sizes grow.
  3. Image manager: set a work cover; library list shows cover thumbnails; article pages show the cover beside the title and in a right rail on wide screens.
  4. Reader-visible version wired through novels (“main version”), paid chapters, and translation sources; cream-theme ranking contrast fixed.

August 4, 2026 (day 130)

Progress

  1. Workspace Fans & community: follower list and counts, mass in-site letters to fans; authors can set Telegram / Discord / QQ in Account settings, shown on work pages and the library author banner.
  2. Account settings: switch workspace appearance (classic dark, warm amber, cream light, parchment). Default stays dark; the same theme applies to workspace, editor, and library.
  3. Fixed near-white callout text on light themes; library cards and filters follow the theme so cream/sepia no longer leaves black boxes.

August 1, 2026 (day 127)

Progress

  1. Mobile lite editor: shows public/draft status; unpublished works link to workspace Publish for first public release; when public, publish or unpublish the current chapter, with a reminder to republish after edits so readers see updates.
  2. Workspace Publish management supports first-time public release (free by default) without finishing full publish setup in the desktop editor first.
  3. Outline tags and editor toolbar actions use high-contrast dark text so labels stay readable on light screens and phones.
  4. Readers can send paper-style in-site letters to authors; authors review inbox and sent mail from the notification bell.

July 31, 2026 (day 126)

Progress

  1. Novel versioning: set a reader-visible version and merge back to main.
  2. Workspace AI chat: streaming replies, automatic retries, and clearer error messages.
  3. Author editor image manager: upload, download, OCR, delete, and copy image links.
  4. Import supports Word (.docx); import methods are swipeable cards for easier mobile use.
  5. On phones, the editor opens a simplified writing mode; use a desktop browser for the full workspace.

May 15, 2026 (day 49)

Progress

  1. AI chat now supports both OpenAI chat/completions and Anthropic provider protocols so the in-browser chat works with either setup.
  2. Fixed formatting on the Privacy Policy and Terms of Service pages.

May 14, 2026 (day 48)

Progress

  1. Added Privacy Policy and Terms of Service pages to the official website.
  2. Shipped 100% disk-persisted fields for Chinese/English/Japanese/Korean with no real-time AI translation.
  3. Expanded development progress into a maintainable standalone data format (JSON/TS), so adding a new day no longer requires component logic changes.

May 13, 2026 (day 47)

Progress

  1. Moved layout strategy from the publish modal into the AI Writing area.
  2. Fixed common AI reflow errors and improved .env.production fallback path resolution.
  3. Bug fix: importing AI chat JSON no longer overwrites previous conversations.

May 10, 2026 (day 44)

Progress

  1. Favorite sentence selection implemented via in-body text selection trigger.
  2. Write-thought popup implemented with wallet check and length limit.
  3. One-click switch between all highlights and original text implemented via tabs.
  4. Bug fix: chapter association list in video management now loads actual chapters.
  5. Bug fix: white text on white background in create-novel modal resolved.
  6. Bug fix: annotation popup no longer disappears immediately after selection.

May 5, 2026 (day 39)

Progress

  1. Integrated Stripe payments, but end-to-end payment collection has not been tested yet.
  2. Improved the visual design of the author workspace.
  3. Added a copy-message button to AI chat.

April 28, 2026 (day 31)

Progress

  1. Bug fix: reader page no longer jumps back to chapter 1 after a refresh while reading chapter 3.
  2. Shared helpers to extract character appearances per chapter; “one-click whole book” batch extraction added on the large-format reader page and the author outline sidebar.
  3. Account settings: choose/switch the model used for translation and related tasks.
  4. Bug fix: multilingual translation could translate chapter 2 when chapter 3 was selected—fixed by removing incorrect fallbacks to draft / full-book text when translating a single chapter.

April 27, 2026 (day 30)

Progress

  1. Switched the site from HTTP to HTTPS.
  2. Workspace PDF signature: choose a PDF and a PNG signature; drag and resize the box on the preview, then download a merged PDF. Processing stays entirely in the browser—nothing is uploaded to the server.
  3. Landing and pitch-deck branding: renamed the Chinese name from 郴郴文库 to 巴别塔; English site label updated to babel towel (domain babeltowel.com).

April 26, 2026 (day 29)

Progress

  1. Renamed the workspace “active wallets” dashboard to “active users”: counts now dedupe by author identity across email/password and MetaMask sign-in (DAU/WAU/MAU, trend chart, and event breakdown copy + API fields updated accordingly).
  2. Workspace AI chat: conversations persist locally in the browser across visits; you can start a new thread, delete the current thread, or clear all saved threads.

April 25, 2026 (day 28)

Progress

  1. Added workspace AI chat via server-side OpenAI-compatible chat/completions; default model is claude-opus-4-6 (overridable with CLAUDE_MODEL).

April 22, 2026 (day 25)

Progress

  1. Workspace video tools: WAV uploads are now supported (same as other non-MP3 sources—transcoded to MP3 for storage).
  2. Fixed upload failures for large files / certain environments (multipart parsing, timeouts, etc.).
  3. Upload flow split: quick upload + persist first, then background transcoding with a browser alert when finished; list shows a “transcoding” state while processing.

April 17, 2026 (day 20)

Progress

  1. Email + password sign-in.
  2. Admin tools (ADMIN_ADDRESS): manually grant, extend, or revoke paid (VIP) membership from the workspace.
  3. When the user has not chosen a UI language and is not connected with a wallet, default the UI language from IP geolocation (e.g. simplified Chinese for mainland China, traditional Chinese for Hong Kong / Taiwan).

April 17, 2026 (day 19)

Progress

  1. Author workspace → Multilingual translation: overview of saved per-language drafts by chapter, language switcher for read-only previews, and links to the library in each language; added chapter-preview API.
  2. Publish settings → “AI auto layout (keep images)”: optional author instructions (prompt) saved on the publish record and passed into the background reflow worker / DeepSeek prompt (core safety rules win on conflict; length capped around 2000 chars).
  3. Landing timeline updated: https://babeltower.hk/

April 14, 2026 (day 16)

Progress

  1. Delete extracted MP3: added a server API (remove from index and delete the audio file) and a delete button next to “Open link” and “Attach to chapter”.
  2. Speech-to-text is now supported.
  3. Workspace (video tools): https://babeltower.hk/workspace

April 13, 2026 (day 15)

Progress

  1. Workspace video tools: upload MP4, server extracts MP3, authors can attach audio to a book chapter; readers hear it on the chapter “Listen” tab.
  2. Fixed author-hosted narration MP3 seeking/scrubbing in the reader: audio-host now supports Accept-Ranges and HTTP 206 byte-range responses.
  3. Startup roundtable feedback (summary): positioning vs benchmarks; clarify target users and creator-first onboarding; communicate highlights in plain, punchy language.
  4. Same session: mainland users may bounce if the landing page is English-only—consider a zh/en toggle on the home page; “Not secure” browser warnings also hurt trust—prioritize HTTPS and certificates.
  5. More notes: moderation / sensitive content expectations; make technical messaging accessible when hiring growth/ops; explore richer reader interactions beyond end-of-chapter comments (e.g. inline highlights, term/allusion explainers, short clips) with compliance in mind.
  6. Product shape questions: mobile app plans; licensing classic works for a global audience; piracy education for hobby authors and possible paid anti-theft services for higher-earning creators.

April 10, 2026 (day 12)

Progress

  1. Paid vs free: all author-workspace / editor flows that ultimately call an LLM or the Python AI service are gated to authors with an active subscription; unpaid authors keep rule-based chapter splitting, while AI chapter split requires membership.
  2. Aura novel-evaluation is integrated: VIP authors get 1 evaluation/day (Asia/Shanghai); A/S grades unlock free platform English translation chars; quality board under Library → 作品 · 质量.
  3. Web app source lives under `apps/web` in the repo (example deploy path: `/root/chenchen-lib/apps/web`).

April 9, 2026 (day 11)

Progress

  1. Added chapter comments: users can comment after connecting wallet
  2. Updated frontend audio UX: when server MP3 cache exists, default to "Play Cached MP3" instead of "Generate MP3"
  3. Added per-chapter word count display
  4. Reader markdown: long code/quoted blocks now auto-wrap while preserving readability

April 8, 2026 (day 10)

Progress

  1. Fixed bug "Outline sync timed out, please check your network and retry." with backend sanitization + frontend no longer sending chapter body in this flow
  2. Fixed bug "Clicking add chapter also creates a volume." so add chapter now creates chapter only
  3. Added Arabic TTS support

April 5, 2026 (day 7)

Progress

  1. Improved bulk .txt import
  2. Frontend: added intro / pitch navigation page
  3. Homepage refresh

April 4, 2026 (day 6)

Progress

  1. Audiobook uploads
  2. Edit audiobook title & details
  3. Reader audiobook playback

April 3, 2026 (day 5)

Progress

  1. Tickets: new page
  2. Human polish after AI generation
  3. Long published content no longer breaks layout
  4. Translated reading keeps image positions (no whole-chapter image loss)
  5. Reduced Chinese leftovers in English mode
  6. Volume / section delete fixes
  7. Scalability note: architecture for ~60k novels
  8. Migration plan
  9. Capacity estimate: 10k authors & free catalog
  10. Ticket images support

April 2, 2026 (day 4)

Progress

  1. Single source of truth for titles
  2. Removed default AI character sheet
  3. Publish using rendered format
  4. Plot outline management
  5. Auto-save current chapter to disk
  6. Chapter content ordering fixes
  7. Duplicate chapter fixes
  8. Loading-state bug fixes
  9. AI layout bug fixes
  10. Layered publish strategy
  11. Copy + AI tagging & synopsis
  12. Social sharing for works
  13. One-click Douyin-style video (frontend)
  14. Multilingual translation
  15. Mixed zh/en translation bug fixes
  16. Partitioned translated content

April 1, 2026 (day 3)

Features

  1. Image library
  2. Chapter TOC
  3. DeepSeek-assisted read, layout & chapter split import
  4. True title display bug fix
  5. Publish all chapters in one action
  6. Hide chapter TOC
  7. Visited link styling
  8. Remember read state after refresh
  9. Markdown editor
  10. Read-only published chapters
  11. Reader typography
  12. Author/reader view sync fixes
  13. Reader social sharing
  14. Reader reading mode
  15. AI layout tooling
  16. Navigation / anchor bug fixes

March 31, 2026 (day 2)

Features

  1. Connection failure help modal
  2. Frontend layout tweaks
  3. Next-chapter tab
  4. Billing / finance
  5. Removed static payment QR
  6. Novel sharing
  7. Reader access controls
  8. Reading experience bug fixes

March 30, 2026 (day 1)

Shipped

  1. Frontend scaffold
  2. miroFish backend
  3. Wallet connect & role detection
  4. Character management
  5. Resume from checkpoint
  6. Card-style outline
  7. Creator / reader split
  8. Author dashboard
  9. Publishing