Installation Guide

May 26, 2026 · View on GitHub

Step-by-step guide to install and use My Translator on macOS.

Requirements

  • macOS 13 or later (Apple Silicon — M1/M2/M3/M4)
  • Soniox mode (recommended): Soniox API key (pay-per-use, ~$0.12/hour)
  • OpenAI Realtime mode (premium): OpenAI API key (~$4/hour — much pricier, but returns native translated voice)
  • Local mode: ~5 GB free disk space (for AI models, one-time download)
  • TTS narration (optional, text engines only): See TTS Guide for provider options

Step 1 — Download

Download the latest .dmg from: Releases — macOS

Choose the right file:

  • MyTranslator_x.x.x_aarch64.dmg — Apple Silicon (M1/M2/M3/M4)
  • MyTranslator_x.x.x_x64.dmg — Intel Mac

Step 2 — Install

  1. Open the .dmg file
  2. Drag My Translator into the Applications folder
  3. Eject the DMG

Step 3 — First Launch

Open My Translator from Applications.

✅ The app is signed and notarized — macOS should allow it without any security warnings.

Step 4 — Grant Screen Recording Permission

On first launch, macOS will ask for Screen & System Audio Recording permission:

  1. Click Open System Settings when prompted
  2. Find My Translator in the list
  3. Toggle the switch ON
  4. macOS will ask to Quit & Reopen — click that button

This permission is required for the app to capture system audio (YouTube, Zoom, podcasts, etc.)

Step 5 — Get a Soniox API Key

Soniox provides real-time speech recognition and translation.

  1. Go to console.soniox.com → create an account
  2. Add billing:
    • Click Billing in the left sidebar
    • Add a payment method
    • Add funds ($10 minimum — lasts ~80+ hours at $0.12/hour)
  3. Create API key:
    • Click API Keys in the left sidebar
    • Click Create API Key
    • Copy the key (format: soniox_...)

💡 Soniox charges ~$0.12/hour of audio processed. $10 ≈ 80+ hours of translation.

After pasting the Soniox key in Settings, the Soniox engine becomes the active option:

Settings — Soniox engine selected with API key

Step 5b — Get an OpenAI API Key (Optional)

Skip this step if you only plan to use Soniox or Local mode.

OpenAI Realtime is the premium engine — it returns translated text and translated speech audio over a single stream, so no separate TTS provider is needed. The trade-off is cost: about $4/hour vs Soniox's $0.12/hour (~34× more expensive at provider list rates).

  1. Go to platform.openai.com → create an account
  2. Add billing:
    • Click Settings → Billing
    • Add a payment method and add credits ($10 ≈ ~2.5 hours of translation)
  3. Create API key:
    • Click API keys → Create new secret key
    • Copy the key (format: sk-...)

⚠️ Cost warning: OpenAI Realtime is roughly 34× pricier than Soniox. Use it for high-stakes meetings where translation quality and native voice matter. For general use, Soniox is the better default.

📊 See OpenAI Realtime vs Soniox benchmark for a real-world comparison.

After pasting the OpenAI key in Settings, the OpenAI Realtime engine becomes selectable:

Settings — OpenAI Realtime engine selected with API key

Step 5c — Get a Qwen LiveTranslate Flash API Key (Optional, free preview)

Qwen LiveTranslate Flash (Alibaba DashScope) is the free preview engine — fastest of the three (~4 s first token), supports 60+ languages, returns translated text only (no native voice, no source-transcript panel).

⚠️ IMPORTANT — must pick Singapore region. The app connects to the international endpoint dashscope-intl.aliyuncs.com. Keys created in other regions (China Beijing, Hong Kong, US Virginia, Germany Frankfurt) are rejected and the app raises WebSocket error the moment you press Start.

  1. Open https://bailian.console.alibabacloud.com (Alibaba Cloud Model Studio).
  2. Before sign-in / sign-up, click the region dropdown in the top-right and pick Singapore. If you've already signed in to another region, switch to Singapore — you may need to register a separate workspace for this region.
  3. Once inside the Console (top-right still shows "Singapore"), activate the Model Studio (DashScope) service if prompted.
  4. Go to API Keys, click Create API Key, name it anything.
  5. Copy the key immediately — it's only shown in full once.
  6. In Settings → pick the Qwen LiveTranslate Flash engine → paste the key.

Settings — Qwen LiveTranslate Flash engine selected with API key

Qwen Live Flash notes:

  • Must pick a source language before Start. Unlike Soniox/OpenAI which auto-detect, Qwen Live needs the source language up front — the source picker automatically hides "Auto-detect" when this engine is selected.
  • No dual-panel view (the model returns translation only, no source transcript). Translation-only display.
  • No native voice output / custom TTS — avoids the speaker → mic feedback loop.
  • Currently in free preview. Pricing may change once it leaves preview — watch Alibaba Cloud announcements.

Troubleshooting WebSocket error with Qwen

SymptomCommon causeFix
WebSocket error immediately on StartKey created in a non-Singapore regionRecreate the key in Singapore (see step 2 above)
Error after ~5–10 secondsRight region but Qwen Live model not enabledModel Studio → Model Square → enable qwen3-livetranslate-flash-realtime
Translates one sentence then stallsSource language left on "auto"Settings → Source language → pick the actual language (e.g. Japanese)
Flaky errorsNetwork blocks dashscope-intl.aliyuncs.com (corp firewall / VPN)Try a different network (4G/5G) or disable VPN

Step 6 — Configure the App

  1. Click ⚙️ (or press ⌘ ,) to open Settings
  2. Go to the General tab
  3. Paste your Soniox API key and/or OpenAI API key (whichever engines you want enabled)
    • A green dot ✓ next to each key field means the key format looks valid; click Test to ping the provider live
    • Engines without a valid key are greyed out in the dropdown
  4. Choose translation type:
    • One-way: Select Source language and Target language
    • Two-way: Select Language A and Language B (for bilingual meetings — the app auto-detects and translates both directions). Two-way is unavailable on OpenAI Realtime — use Soniox or Local for two-way.
  5. Choose Translation Engine:
ModeSpeedQualityCostVoice outputSource transcriptInternet
☁️ Soniox~2 s9/10~$0.12/hrVia TTS (free–$8/hr)✅ Yes (dual panel)Required
⚡ OpenAI Realtime~1.5 s9.5/10, very idiomatic~$4/hrOff by default✅ Yes (dual panel)Required
🌏 Qwen LiveTranslate Flash~4 s8/10, 60+ languagesFree (preview)❌ None❌ None (translation only)Required
🖥️ Local MLX~10 s7/10FreeVia TTS✅ YesNot needed
  1. Click Save & Close

Local MLX requires Apple Silicon (M1+) and ~5 GB disk. Models are downloaded automatically on first use.

OpenAI Realtime supports 13 target languages: en, es, pt, fr, de, it, ru, hi, id, vi, ja, ko, zh. For Thai or other languages, use Soniox. The custom TTS toggle is automatically disabled while OpenAI Realtime is selected (audio comes from the model itself).

Step 7 — Enable TTS Narration (Optional)

Want translations read aloud? Three TTS providers are available:

ProviderCostQualitySetup
🎙️ Edge TTSFreeNaturalNone
🌐 Google Chirp 3 HDFree 1M chars/moNear-humanGoogle Cloud API key
✨ ElevenLabs~$5/mo+PremiumElevenLabs API key

Quick setup (Edge TTS — free):

  1. Settings → TTS tab → Provider: Edge TTS
  2. Choose a voice → Save & Close
  3. On main screen, click the TTS button (or ⌘ T) to enable

For Google or ElevenLabs:

See TTS Guide for step-by-step API key instructions.

Step 8 — Start Translating!

  1. Go back to the main screen
  2. Click ▶ (or press ⌘ Enter) to start
  3. Play any audio on your Mac (YouTube, Zoom, podcasts...)
  4. Translations appear in real-time!

View modes:

  • Single (default): Translation text only
  • Dual: Source | Translation side-by-side (toggle with panel button, bottom-right)

Font size: Use A-/A+ buttons (bottom-right on hover) to adjust

Choosing the translation mode

If you have both a Soniox and an OpenAI key configured, the first time you start a session the app asks which engine to use:

Choose translation mode — Standard vs OpenAI Realtime

You can switch any time from the engine pill in the toolbar.

Dual-panel view with OpenAI Realtime

In Dual view the source transcript appears on the left and the translated text on the right — OpenAI's whisper transcription and translated output stream side-by-side:

Dual-panel translation running with OpenAI Realtime

Keyboard Shortcuts

ShortcutAction
⌘ EnterStart / Stop
⌘ ,Open Settings
EscClose Settings
⌘ 1Switch to System Audio
⌘ 2Switch to Microphone
⌘ TToggle TTS narration

Troubleshooting

No translation text appears

→ Check Screen & System Audio Recording is enabled in System Settings (see Step 4)

"No API key" error

→ Open Settings (⚙️) and paste a Soniox key (Step 5) and/or an OpenAI key (Step 5b) for whichever engine you selected

OpenAI Realtime: engine option is greyed out

→ The OpenAI key field is empty or the key format is invalid (must start with sk-). Paste a fresh key, then click Test to verify

OpenAI Realtime: "Two-way" toggle is hidden

→ This is expected. Two-way mode is only available on Soniox and Local MLX. Switch engines if you need it

Translation cost is much higher than expected

→ Confirm which engine you're using. OpenAI Realtime is ~$4/hour vs Soniox's ~$0.12/hour. The engine indicator shows under the dropdown in Settings

"No microphone found" error

→ Mac Mini has no built-in microphone. Connect an external mic (USB, headset, AirPods)

TTS not working

→ See TTS Guide — Troubleshooting

Updating

My Translator includes auto-update. When a new version is available:

  1. A green badge appears on the ⚙️ settings icon
  2. Open Settings → About tab → click Download & Install
  3. The app will restart automatically with the new version

No need to download DMG files manually for future updates!