EchoForge v2.5.1 — Button Bug Fix

v2.5.1

🐛 Bug Fix

Root cause: bindSettings() referenced $("btn-test") and $("test-result") — elements that no longer exist in v2.5 HTML (replaced by btn-test-fish / fish-test-result in bindVoiceEngines). This null reference threw an exception, aborting bindSettings() before $("btn-save-settings") could be bound — cascading failure prevented ALL subsequent boot steps (LLM, engines, libs, scripts, generate, agent binding).

Impact: Every save/test button was dead. Settings could not be saved. No tab navigation after initial load.

Fix: Removed the stale 8-line handler block. Also fixed ms_enabled save expression.

Verified: 12+ buttons all bound, script parse works, project modal works, offline generation works end-to-end.

📦 Download

EchoForge-v2.5.1-win-x64.exe (~55 MB)

EchoForge v2.5.0 — Performance + Project Modal + UI Fix

v2.5.0

⚡ Performance Fix

Removed 15 backdrop-filter blur instances from cards (the #1 cause of UI freezing). Cards now use solid white — butter-smooth on all machines.

📁 Project Management UI

Replaced the s2/r1/d1 command-line prompt with a proper modal panel: project list with switch / rename / delete buttons + new project input field.

🔧 UI Fixes

• Sidebar no longer disappears on settings page (fixed unclosed <div>)
• HTML structure fully balanced (140 open = 140 close)
• Fish label: removed 4模型 suffix
• boot rejection fixed (stale DOM refs)

📦 Download

EchoForge-v2.5.0-win-x64.exe (~55 MB)

EchoForge v2.4.0 — Parallel Engines · Projects · Password · AV Fix

v2.4.0

🎙️ Parallel TTS Engines (no more switching)

All three engines are independent and off by default:

Engine	How to enable
Fish Audio	paste API key → enabled
MiniMax	paste key (Coding Plan sk-cp works) → enabled
Microsoft Edge	flip the toggle (no key needed)

Enable any combination — a default engine picker auto-lists only the enabled ones, and per-character channel overrides keep working.

📁 Projects Workbench

Each production lives in its own project — scripts, characters, voice bindings and Agent series are fully isolated. Create / switch / rename / delete from the sidebar. No more crosstalk between dramas.

🔐 Access Password (opt-in)

Set a password in Settings (SHA256-stored). All API calls then require a token, with a lock-screen on open. Recommended whenever you serve over LAN (--host 0.0.0.0). Off by default.

🛡️ Windows Antivirus Mitigations

The EXE now embeds a proper application manifest (asInvoker, per-monitor DPI, Win10/11 compatibility) and is compiled with ASLR / DEP / High-entropy VA / Terminal-Server-Aware flags — measurably fewer heuristic flags. For SmartScreen: More info → Run anyway (documented in README).

📦 Download

EchoForge-v2.4.0-win-x64.exe (~55 MB)

EchoForge v2.3.0 — Multi-Channel Casting · Manual Voiceover

v2.3.0

🎭 Mix TTS Channels in One Drama

Each character can now use a different provider:

Character	Channel
Narrator	Fish Audio (native [emotion] brackets)
Hero	MiniMax (Coding Plan supported)
Villain	Microsoft Edge (opt-in)
Guest	🎙 Your own recording

• Per-character override with global fallback
• Microsoft is opt-in: disabled by default; enable it in Settings, otherwise MS-assigned characters fall back to the global provider
• Only Fish parses [bracket] emotions natively; MiniMax/Edge auto-convert (strip + parameter mapping)

🎙 Manual Voiceover (Human Dubbing)

• Set a character to 🎙 Manual mode, then upload a recording for each line right in the segment table
• Post-production DSP: change voice preset/pitch on recordings anytime (re-render applies transforms)
• Manual lines are mixed with TTS lines seamlessly, with all BGM/SFX ducking intact

📦 Download

EchoForge-v2.3.0-win-x64.exe (~55 MB)

EchoForge v2.2.0 — Voice Pool · Emotion Compat · Frosted Glass

v2.2.0

🗑 Voice Pool Management

• Delete any voice from the pool (fish has no slot limit — voice IDs are unlimited)
• Manual voices are removed directly; API-fetched voices are hidden via blacklist
• ↺ Restore deleted brings everything back

🔄 Provider Switching Fixed

• Switching Fish / MiniMax / Microsoft now instantly previews that provider's voices — no more guessing how to fetch
• Saving voice config auto-loads the voice list into Cast & Voices

🗣️ Emotion Bracket Compatibility

[softly]-style markers behave correctly on every provider now:

Provider	Behavior
Fish Audio (S2)	kept natively (full expressiveness)
MiniMax	stripped + mapped to emotion parameter (verified: reading time back to normal)
Microsoft Edge	stripped + mapped to volume/rate (edge-tts used to read tags aloud — fixed)

✨ Frosted-Glass UI

Complete redesign: translucent white cards with backdrop blur, restrained single-accent palette, refined typography and shadows. Reviewed as production-grade minimal design.

📦 Download

EchoForge-v2.2.0-win-x64.exe below (~55 MB, single file).

EchoForge v2.1.0 — Now on Linux, macOS & Android 🎉

🎙️ EchoForge v2.1.0 — Full Cross-Platform Release

EchoForge now runs on every major platform:

Platform	Download	Launch
🪟 Windows	EchoForge-v2.0.0-win-x64.exe (below, unchanged)	double-click
🐧 Linux	EchoForge-v2.1.0-x86_64.AppImage or EchoForge-linux-x64-v2.1.0	chmod +x && ./EchoForge*
🍎 macOS	EchoForge-macOS-Linux-source-v2.1.0.zip	unzip → bash install-macos.sh → get a real EchoForge.app
📱 Android (native app)	EchoForge-v2.1.0-android-native-arm64.apk	download & install — no Termux, no Python, fully offline UI
🤖 Android (Termux)	EchoForge-Android-Termux-v2.1.0.zip	Termux → bash install-android.sh
📱 Any phone (PWA)	nothing to install	run on a PC with --host 0.0.0.0, phone browser → Add to Home Screen

✨ What is new

• 📱 Native Android app (NEW) — a true native Android rewrite in Java: zero WebView, zero bundled Python, 14 MB vs the old 116 MB shell. Direct Edge-TTS WebSocket (free voices, no key), Fish Audio & MiniMax, native MediaCodec audio pipeline, WSOLA DSP voice presets, LAME MP3 export with ID3 tags, foreground-service rendering with notification, 5-tab material-style UI (剧本/角色/设置/生成/作品), script sharing via FileProvider. Android 7.0+ arm64
• 🐧 Linux single-file binary + AppImage — built and smoke-tested on real Linux (TTS + PWA endpoints verified live)
• 🍎 macOS one-click installer — bash install-macos.sh creates a genuine EchoForge.app bundle you can drop into /Applications, plus an EchoForge.command launcher
• 🤖 Android (Termux) installer — installs all dependencies, creates a global echoforge command, optional boot autostart via Termux:Boot
• 📱 PWA support — installable to any phone home screen with app icon, fullscreen standalone mode, shortcuts
• 📱 Responsive mobile UI — top horizontal scrolling nav, larger touch targets, fullscreen dialogs, full-width mini-player; verified at 390px viewport

📱 The three ways to use EchoForge on a phone

1. Native app (recommended) — install the APK above; everything runs on-device with a real mobile UI
2. Termux (full phone-only) — install Termux from F-Droid, run the installer, everything runs on the phone itself
3. PWA (recommended) — keep EchoForge running on any computer (python app.py --host 0.0.0.0 --port 8080), open http://<pc-ip>:8080 on the phone, tap Add to Home Screen — you get a fullscreen app icon that works over Wi-Fi

All features are identical on every platform: 3 TTS providers (Fish Audio / MiniMax Coding Plan / free Microsoft Edge), DSP voice transformation, SSML breaks, EN/ZH/JA UI, waveform timeline, Agent mode with batch export.

EchoForge v2.0.0 — Multi-Provider TTS · Voice Transform · i18n

🎙️ EchoForge v2.0.0

The biggest release yet — one voice now plays an entire cast.

✨ New in v2.0

🎭 Three TTS Providers

Provider	Key needed	Voices
Fish Audio	fish.audio key	your cloned voices (4 models incl. free tier)
MiniMax	sk-cp-… Coding Plan key works directly	13 built-in Chinese voices
Microsoft Edge TTS	❌ none — free	63 voices (zh/en/ja)

🧬 Voice Transformation (DSP)

Make ONE voice play MANY characters — 12 presets (Boy 少年 / Elder 老者 / Elf 精灵 / Robot 机器人 / Narrator 旁白…) plus manual controls: pitch ±12 semitones (WSOLA time-preserving), formant shift, brightness, breathiness, roughness.

🗣️ SSML Pronunciation Control

<break time="500ms"/> inside any line — works on all three providers (native on Microsoft, auto-converted on Fish/MiniMax).

🌐 UI Localization

EN · 简体中文 · 日本語 switcher in the sidebar.

🌊 Waveform Timeline

Canvas waveform with segment color-layers (SFX in blue) + click-to-seek.

📦 Batch Export + ID3

Agent multi-episode batch export → MP3s with ID3v2.3 tags (title/artist/album/track) → single ZIP. Regular export is now ID3-tagged too.

📱 Android / macOS / Linux

• Android (Termux): bash run-android.sh → phone serves on :8080
• macOS/Linux: python app.py --host 0.0.0.0 for LAN access

📦 Download

File	Description
EchoForge-v2.0.0-win-x64.exe	Single-file Windows 10/11 x64 (~55 MB)

Docs: English · 简体中文 · 日本語

EchoForge v1.3.3 — First Public Release

🎙️ EchoForge v1.3.3

AI-Powered Radio Drama Production Studio — first public release.

✨ Highlights

• 🎭 Fish Audio multi-speaker TTS — 4 models incl. free tier, per-character voice binding, content-hash caching (never billed twice)
• 🎵 Adaptive BGM ducking — [intro ➝ speech(ducked to 25%) ➝ outro] with per-directive fades #BGM=name,in,out
• 🔊 20 built-in SFX (#SFX=name,volume) — thunder, footsteps, birds, campfire… all synthesized, royalty-free
• 🎵 Timed interludes #MUSIC=track,seconds and ending theme #END=track,seconds with fade-out
• 🎹 5 piano-led BGM tracks ≥ 2 min built in + batch upload for your own library
• 🤖 AI Screenwriting — any OpenAI-compatible API + Claude native, one-click model discovery, full project context in every prompt
• 🚀 Agent Mode — auto prologue + episodes, multi-round human feedback, series continuation, auto voice-binding + auto generation
• 📊 Live progress — percentage, stage pipeline, current-line card

📦 Download

File	Description
EchoForge-v1.3.3-win-x64.exe	Single-file Windows 10/11 x64 (~54 MB) — double-click and go

Or run from source: pip install -r requirements.txt && python app.py

📚 Documentation

English · 简体中文 · 日本語

🔑 Setup

• TTS: free API key at fish.audio (model s2.1-pro-free)
• LLM (optional): OpenRouter / DeepSeek / Ollama … see docs/llm-providers.md

SmartScreen may warn (unsigned binary): More info → Run anyway. Upgrades are automatic — just launch the newer EXE.