=== Replica 11 - Script to Speech === Contributors: petrchroustovsky Author: Petr Chroustovsky / d3arts.cz Author URI: https://d3arts.cz Product URI: https://d3arts.cz/obchod/umela-inteligence/replica-11-script-to-speech-elevenlabs-ai-voices/ Tags: replica 11, script to speech, elevenlabs, chatgpt, audiobook, dialogue, voice generation, windows, macos Requires OS: Windows 10/11 (64-bit) or macOS 12+ (Apple Silicon) Stable tag: 1.1.7 License: Proprietary commercial license (d3arts.cz) License file: LICENSE.txt (mandatory attachment in Replica11.zip) License URI: https://d3arts.cz/obchod/umela-inteligence/replica-11-script-to-speech-elevenlabs-ai-voices/ Desktop app (Windows and macOS Apple Silicon) that turns written scripts into structured dialogue data and ElevenLabs spoken audio. == Application Information == * Application name: Replica 11 – Script to Speech * Developer: Petr Chroustovsky * Publisher / website: d3arts.cz * Current version: 1.1.7 (from package.json) * Platform: Windows and macOS Apple Silicon (Electron) * License type: Commercial lifetime license sold via d3arts.cz (up to 2 activation slots) * License document: LICENSE.txt (EULA effective version 1.0 - 8 October 2026) * Trial: 3 days from first launch without an activated license * Product page: https://d3arts.cz/obchod/umela-inteligence/replica-11-script-to-speech-elevenlabs-ai-voices/ * FAQ / guide: https://d3arts.cz/umela-inteligence/from-script-to-audiobook/ * In-app manual: Help → Documentation & Manual (F1) * Release zip must include: Setup installer + readme.txt + LICENSE.txt (not version.txt) Important: A Replica 11 license covers the desktop application only. It does not include OpenAI (ChatGPT) or ElevenLabs credits, subscriptions, or usage fees. You need your own accounts with those services. API costs are billed by them separately. == Description == = Short description = Replica 11 is a Windows and macOS (Apple Silicon) desktop tool that prepares long scripts for ElevenLabs: it detects scenes, sections, characters, dialogue, narration and stage directions, lets you assign an ElevenLabs voice_id to every character, then exports structured data or generates spoken audio locally. = Full description = Replica 11 is the production layer between a written script and AI voices. Paste or load a screenplay-style script. Analyze it locally (no cloud required), or optionally process it with ChatGPT for cleanup, structure and emotional audio tags. Map each character to an ElevenLabs voice. Export to TXT, JSON (Text to Dialogue), Studio JSON, CSV or Markdown — or generate audio in batches via the ElevenLabs Text to Dialogue API straight to a folder on your PC. Nothing is sent without your consent: text goes to OpenAI only after you confirm, and to ElevenLabs only when you press Generate with ElevenLabs. API keys stay on your computer (Electron safeStorage) and are never written into project files or exports. Typical users: screenwriters, podcasters, game developers, audiobook creators, and AI filmmakers who need consistent multi-character dialogue audio. Pipeline: Raw script → Clean text (optional, local) → Analyze locally OR Process with ChatGPT (+ enhance) → Review & Approve structure (ChatGPT path) → Map ElevenLabs voice IDs → Export and/or Generate with ElevenLabs → Local WAV / MP3 files == Features == * Local script parser (offline; no API key required) * Optional ChatGPT / OpenAI structuring and text enhancement * Translate transcription via ChatGPT (full script, selected script text, or selected replica text) * Clean text (local removal of separators, bullets and formatting noise) * Character detection and per-character ElevenLabs voice_id mapping * Character voice memory across analyses / projects * Stability presets (Creative / Natural / Robust) and Similarity (Low / Medium / High) * Merge characters (combine aliases / typos into one voice) * Editable dialogue table (scene, character, voice, text, direction, language, skip) * Direction / emotion audio tags (e.g. [excited], [curious]) * Parentheses-to-audio-tag conversion for common stage notes * Secondary-language tags inside English lines (accent_tags or plain) * Language override for ElevenLabs language_code (auto / en / cs / sk / off) * Narrator reads unroled lines option (scene titles, sections, stage directions) * Batch audio generation via ElevenLabs Text to Dialogue API * Output quality: HQ WAV 48 kHz or LQ MP3 128 kbps * Export: TXT, JSON (Text to Dialogue), Studio JSON, CSV, Markdown * Save / open / new project files (*.replica-11-project.json); API keys never stored in projects * Load TXT scripts * Built-in license activation, revalidation and deactivation * 3-day trial from first launch * Automatic and manual update checks (opens Replica11.zip download in the browser) * Multilingual UI (27 interface languages) * 29 secondary languages for tagging / pronunciation workflows (plus English as base) * In-app documentation (English / Czech HTML manuals) * Local-only encrypted storage of API keys == System Requirements == * Operating system: Windows 10 (64-bit) or Windows 11 (64-bit) * Not supported: Windows 7 / 8 / 8.1, macOS, Linux, mobile * Architecture: 64-bit (x64) * Internet connection: required for license checks, updates, ElevenLabs generation, and optional OpenAI features * Offline-capable features: local Clean text, Analyze locally, editing, and all export formats * Third-party accounts (optional depending on workflow): - ElevenLabs API key for voice generation - OpenAI API key for ChatGPT processing and Translate transcription == Installation == 1. Download Replica 11 from the d3arts.cz product page or the published Replica11.zip package. The zip must contain the Setup installer together with readme.txt and LICENSE.txt. 2. Read LICENSE.txt (End User License Agreement) before installing or activating. 3. Run the Windows installer (Replica 11 Setup *.exe) or unpack / run the packaged application as provided. 4. Launch Replica 11. 5. On first launch a 3-day trial starts automatically if no license is activated. 6. Open Settings (top-right toolbar). 7. Activate your license key when ready (Help / Settings → License), or continue in trial. 8. (Recommended) Add an ElevenLabs API key → Save key → Test API key. 9. (Optional) Add an OpenAI API key for ChatGPT features → Save key → Test API key. 10. Click Save settings. 11. Paste or Load TXT, Analyze locally, assign Voice IDs, then Export or Generate with ElevenLabs. Account / license management: https://d3arts.cz/muj-ucet/?d3_account=en == Configuration == Open Settings from the top toolbar. = App language = Switches the user interface language. Does not by itself change ElevenLabs generation language. = Secondary language = Defines paired language tags inside English lines (for example japanese / czech tags) and the second “set language for all rows” button in the Dialogue table. Default: Czech. = Project shortcuts in Settings = Open project / Save project / Load TXT (also available when license blocking would otherwise limit the app). = License = Activate, Check now (revalidate), or Deactivate on this computer (frees an activation slot). = ElevenLabs settings = * API key (stored locally only) * Default model (free-text model id; default eleven_v4) * Output – quality / format: HQ WAV 48 kHz or LQ MP3 128 kbps * Default narrator voice ID * Language tag conversion: accent_tags (recommended) or plain * Language override: auto / English / Czech / Slovak / off * Characters & voice IDs table (Voice ID, description, Stability, Similarity) = OpenAI settings = * API key (stored locally only) * OpenAI model (curated list; default gpt-6.1-sol) * Processing mode: - local_parser_only - openai_normalize - local_parser_then_openai_cleanup (default) * Text enhancement: off / tags_only / tags_and_interjections * Mark secondary-language names/cities/institutions in English text * Let ChatGPT remove human separators / formatting noise = Save = * Save key stores an API key immediately. * Save settings persists models, quality, accents, enhancement, languages and related options. == ElevenLabs Integration == = API key = Settings → ElevenLabs API → Get Keys (https://elevenlabs.io/app/developers/api-keys) Paste the key → Save key → Test API key. Without an ElevenLabs key you can still analyze and export. Generate with ElevenLabs will not run. = Voice IDs = 1. In ElevenLabs open Voices → My Voices. 2. Open the voice menu → Copy voice ID. 3. Paste the ID onto the matching character in Replica 11. One character = one consistent voice across the project. A default narrator voice ID covers narrator lines when no dedicated ID is set. = Generation behaviour = * Uses the configured Default model (for example eleven_v4). * Sends Stability / Similarity presets with generation. * Applies language tag conversion and language override as configured. * Writes audio to Documents\Replica 11\exports\audio by default (custom folder supported). = Batch limits = Generation splits lines into batches when needed: * at most 2000 characters per batch * at most 10 distinct voice IDs per batch * language change starts a new batch (single-language batches) * stability change starts a new batch A single line is never split; order is preserved. The Generation log shows each batch result and API errors. == Script Processing == = Local analysis = Analyze locally parses the script without OpenAI. It detects: * Scenes — entire line in UPPERCASE * Narration — lines starting with "Did you know?" / "Věděli jste?" * Dialogue — Name: text * Sections — short title lines without a colon * Stage directions — remaining lines It also suggests English voice descriptions and direction tags from context. Narrator and the Czech Vypravěč are treated as the same person. = Clean text = Local cleanup removes separator-only lines, stray punctuation artifacts, bullets/numbering, inline separators and redundant blank lines. Fixes doubled colons after speaker names. = Process with ChatGPT = Optional. Requires an OpenAI API key and user confirmation before sending. * Structures / normalizes the script according to Processing mode * Optionally adds emotion tags and interjections (Text enhancement) * Opens an OpenAI Review window; results apply only after Approve * Existing voice_id values, descriptions and stability are preserved across runs = Translate transcription = Optional ChatGPT tool (File → Translate transcription… or the Script toolbar). * Choose a target language * Optionally keep secondary-language words * Optionally translate only the selected text * A selection inside a dialogue row is translated in that row. The script window stays put and the script is not analyzed again * Translating the whole script still replaces the Script panel (view jumps to the start). Replicas update only after you analyze and approve the structure again * Long scripts are sent in chunks and reassembled * Requires OpenAI; does not call ElevenLabs = Dialogue editing = After analysis you can filter by scene / character / language, set language for all rows, add/delete lines, skip rows, undo recent edits, and merge characters. == Audio Generation == 1. Assign Voice IDs (and optional Stability / Similarity). 2. Choose HQ or LQ in Settings. 3. Pick or confirm the audio output folder. 4. Click Generate with ElevenLabs. 5. Watch the Generation log for batch progress and errors. 6. Open the export folder to collect WAV or MP3 files. Supported exports (work without an ElevenLabs key): * TXT — readable transcript by scene * JSON (Text to Dialogue) — inputs array with text + voice_id * Studio JSON — ElevenLabs Studio from_content_json structure * CSV — full line table for review in spreadsheets * Markdown — readable script with headings Lines marked Skip are excluded from export and generation. == Frequently Asked Questions == = Do I need ChatGPT / OpenAI to generate voice audio? = No. Voice generation uses ElevenLabs only. Analyze locally → assign Voice IDs → Generate with ElevenLabs. An OpenAI key is not required for that path. = What works without which key? = * No keys: Load TXT, Clean text, Analyze locally, edit dialogue, export all formats, save/open projects * ElevenLabs only: full voice generation pipeline * OpenAI only: Process with ChatGPT, Translate transcription, optional enhancement * Neither cloud call runs until you explicitly confirm / start it = I only want spoken dialogue. What is the minimum setup? = 1. Activate license (or use the 3-day trial) 2. Add ElevenLabs API key → Save → Test 3. Set model and HQ/LQ if needed 4. Skip OpenAI 5. Analyze locally → paste Voice IDs → Generate = Are API keys stored in the project file? = No. Keys stay on this PC only. Project files can be shared without exposing credentials. After reinstall, enter keys again in Settings. = Does changing the OpenAI model change ElevenLabs voices? = No. OpenAI settings affect ChatGPT text passes only. ElevenLabs Default model, Voice IDs, Stability / Similarity and HQ/LQ control audio. = What is Translate transcription? = A ChatGPT-powered translation of the script text (whole script, a selection in the script, or a selection in a dialogue row). It changes text only; it does not generate audio. A whole-script translation still needs analyze and approve before replicas change. A selection inside a replica is replaced in place. = Where do I find a Voice ID? = ElevenLabs → Voices → My Voices → Copy voice ID → paste onto the character in Replica 11. = Which systems are supported? = Windows 10 and Windows 11 (64-bit) only. No macOS, Linux or mobile builds. = Is there a free trial? = Yes. Replica 11 can be tried for 3 days from first launch. Download is available from the product page; the full commercial license unlocks ongoing use. When the trial ends (or if no valid key is active), Settings → License shows a prominent red alert and the app requires purchase and activation of a license key. = Must I accept the EULA during installation? = Yes. The Windows installer displays LICENSE.txt and installation continues only after you accept the End User License Agreement. == Troubleshooting == = OpenAI error 429 “exceeded your current quota” = Your OpenAI account is out of credit or over its limit. Check billing at platform.openai.com. This is not an application bug. = A line was not generated = The voice_id is missing. Fill it in for the character, or set a default narrator voice ID in Settings. = Generation fails with WAV / HQ output = HQ (WAV 48 kHz) may require a higher ElevenLabs plan. Switch Output to LQ (MP3 128 kbps) in Settings. = Lines detected incorrectly = Check script formatting (Name: text, UPPERCASE scenes), run Clean text, or use Process with ChatGPT. = A character appears multiple times = Use Merge characters below the characters table. = English dialogue sounds like another language = Set Dialogue language to English and Language override to English (or fix per-row languages and leave override on auto). Check language-tag conversion (accent_tags vs plain). = API keys missing after reinstall = Keys are tied to the application user profile. Enter them again in Settings. = fetch failed / certificate errors on this PC = Some antivirus products (for example Avast Web Shield / HTTPS scanning) can interfere with HTTPS. Temporarily disable HTTPS scanning for Replica 11 or add an exception, then retry. = License / trial blocked = Activate a valid license key in Settings, or wait until online revalidation succeeds. Use Deactivate on this computer before moving the license to another machine if you hit activation limits. = Updates = Help → Check for Updates. If a newer version is published, the app opens the Replica11.zip download URL in your browser. Install the new package manually. == Changelog == Keep this section updated when releasing a new version. Do not invent older history. = Version 1.1.7 = * Help → Tutorial opens the YouTube tutorial. Help → Report an Issue opens the contact page on d3arts.cz. The app does not send the script, project, or API keys. * The narrator role name follows the UI language: the unassigned-lines option, the character column for that role, and the default narrator voice in Settings. Voice assignment still treats the narrator as one character. * macOS Apple Silicon DMG can be signed with Developer ID and notarized (`npm run dist:mac:signed`). * macOS update checks read https://d3arts.cz/wp-content/uploads/replica11/mac/version.txt and open Replica11.dmg from the same folder. = Version 1.1.6 = * What's new window opens automatically after each update with a table of changes by version (also available any time via Help → What's New). * English split into English (GB) and English (US): per-row language en-GB / en-US and bulk buttons in the dialogue table. Generated audio gets a [British accent] / [American accent] tag on each line (language_code stays "en"); secondary-language passages switch back to the same accent. Translate transcription offers both variants with British / American spelling and vocabulary. * Settings → ChatGPT → English variant (neutral / GB / US): after analysis English rows get en-GB / en-US, voice descriptions get the accent (e.g. "British accent (Received Pronunciation)"), and ChatGPT enhancement picks British- or American-style audio tags and interjections. * Regional variants work the same way for Spanish (Spain / Mexico / Argentina / Chile / United States), Portuguese (Brazil / new Portugal) and French (France / new Canada): rows get an accent tag such as [Mexican Spanish accent] in generated audio, voice descriptions get the accent (also after "Language for all rows"), translation uses regional vocabulary, and ChatGPT enhancement uses region-specific audio tags and interjections when the variant is the secondary language. * Undo history is configurable in Settings → Editing (10–200 steps, default 50; previously fixed at 5). History lives in memory only (like Photoshop) and covers all dialogue-table edits (typing in one cell = one step), language/Skip changes, analysis, Clean text, Translate and Load TXT. New Redo button and Edit menu shortcuts Ctrl+Z / Ctrl+Y (Ctrl+Shift+Z); inside text fields the native text undo stays. = Version 1.1.5 = * File → New Project (Ctrl+N) clears the workspace after asking whether to save the current project (Save / Don't Save / Cancel). * Translate transcription can replace a selected phrase inside a dialogue row in place — the script window stays put and the script is not analyzed again. Whole-script translation still replaces the Script panel and needs analyze + approve. * In-app manuals (Help → Documentation) updated for File menu, projects and in-place replica translation. = Version 1.1.4 = * Fixed mojibake in the installer EULA page (UTF-8 em dash replaced with ASCII hyphen in LICENSE.txt). * Windows Setup and application binaries digitally signed (Certum Code Signing, CN=Petr Chroustovsky, SHA-256 + Certum timestamp). * File → Open Recent remembers up to 10 last saved/opened projects (Clear Recent File List clears the list). = Version 1.1.3 = * Settings → License shows a prominent red alert when the free trial ends, no key is active, or the license is invalid/expired — purchase and activation required. * Installer EULA acceptance (LICENSE.txt) and mandatory zip attachments (readme.txt + LICENSE.txt) carried forward from 1.1.2. * Documentation/FAQ updated for trial expiry messaging and licence requirements. = Version 1.1.2 = * NSIS installer now shows the proprietary EULA (LICENSE.txt) and requires user acceptance before installation. * Release package Replica11.zip includes Setup installer + readme.txt + LICENSE.txt (version.txt is not packed inside the zip). * Added pack:zip / dist:zip release packaging scripts for consistent public downloads. * Documentation readme.txt updated for the current release workflow and licence attachment. = Version 1.1.1 = * Stable release documented from the local project (package.json version 1.1.1). * Windows desktop Script to Speech workflow: local parse, optional ChatGPT processing, ElevenLabs voice mapping and batch generation. * Translate transcription via ChatGPT (full script, script selection, or in-place replica selection; optional keep secondary-language words). * ElevenLabs defaults centered on eleven_v4 with HQ WAV / LQ MP3 output options. * Multilingual UI and secondary-language tagging for mixed-language scripts. * Local encrypted API key storage; keys excluded from project files and exports. * Commercial licensing with 3-day trial, activation / revalidation / deactivation. * In-app documentation manuals and update check against the published package on d3arts.cz. -- End of readme.txt -- When publishing a new build, add a new "= Version X.Y.Z =" block above older entries and update the Stable tag in the header. )