Interview support. Live translation included.
Follow both sides of a call, see the translation and get short answers based on your CV and the conversation. No bot joins the meeting.
The other person only hears or sees what you choose to send or share.
Five-day demo · Windows 10/11 · cloud access or local models
Three ways to stay in the conversation.
Get help with an interview, practise a language, or keep a companion beside your call.
Find the words.
Keep your own story.
Get a short, speakable prompt from your CV, job description and Q&A bank. A fuller answer follows when you need it.
Your context: CV + job description + Q&A
Follow the meaning.
Find your next reply.
Read the original and translation together. Learn the words and grammar in context, then find a reply you can use.
Both sides of the call. Languages depend on your speech and translation engines.
Stay with the call.
Keep help within reach.
A companion follows the conversation and helps you find a useful response. Keep its panel above your call, or show only the prompts.
Adjust the panel size and opacity.
Try the panel controls
More help, when you need it.
Voice, screen context and recordings work alongside the mode you choose.
Voice outputSay a translated reply in your own voice
Install My Voice from voice settings, then record a 10–60 second sample to create a local voice profile. My Voice requires a compatible NVIDIA GPU. You can also choose local Piper or Microsoft Edge Neural voices. Speech output is optional; add VB-CABLE to send it back into the call.
The local voice sample and profile stay on your PC. Cloud voice services receive the text they speak.
Screen contextBring a task or diagram into your answer
Capture a task, diagram or code with a shortcut. Choose a monitor, window or region, then extract text with local RapidOCR or send the image to a vision model.
Local OCR reads the image on your PC. Extracted text may still go to online AI; a vision provider receives the image itself.
Call historyReview the call instead of taking notes
Search the transcript, play the recording and export the text. Import WAV or MP3 files, transcribe them with an installed local speech model, and request a summary or interview review.
Recording is optional. Summaries and reviews use your selected AI. Deleting a stored call does not delete original imports, exports, backups or copies held by providers.
How your call becomes a prompt
01.1 / SYSTEM OVERVIEW
Talkaside transcribes both sides of the call. Your chosen engines translate the conversation and suggest answers using your CV, job description and other material. Speech back into the call is optional.
Speech to text
Separate tracks for you and the other person.
Your material
CV, job description, Q&A and screen capture add context to the assistant.
Prompts + translation
Interview prompts, language practice or a call companion.
Why it matters. Follow the translation and get an answer from your own context without switching away from the call. Read the prompt yourself, or turn on speech output when you want it.
History stays on your PC. Connected cloud providers process the audio or text they need. Speech into a call requires VB-CABLE and a configured voice. Simplified flow, not a latency measurement.
Set up once. Test before your call.
Start with on-screen prompts. Voice output, cloning and extra providers can wait.
Windows 10 or 11, 64-bit. Follow the activation screen. Check the source and file before running an unfamiliar installer.
Choose cloud providers or supported local models and test them in settings. For example: Deepgram for speech and Groq for answers. Cloud services need your own access; local models need a download or an existing AI server.
Check the microphone and call audio, then try a question before the interview. Add your CV and job description if you want answers based on them.
The demo covers the app licence, not provider usage. Cloud AI, speech recognition and paid voices are billed separately by the services you connect. Local processing uses your own hardware. VB-CABLE is only needed for speech back into the call.
Your calls do not pass through us
Your history stays on your PC. Connected cloud services receive relevant data directly from your computer, not through a Talkaside call server.
Your files stay on your PCLocal history, documents and recordings
- History, recordings, document context and My Voice profiles are stored on your computer. Enabled online features may still send relevant content to their providers.
- Recording, translation, speech output and prompts have separate switches. History controls manage stored calls; original files, exports, backups and provider copies are separate.
You choose what leaves your PCYour providers, your encrypted keys
- Online speech recognition sends audio; translation, AI and speech output send the relevant text or context. Vision may send a captured image. Local OCR can still pass extracted text to online AI. Saved provider keys are protected for your Windows account using DPAPI.
- Web search may send queries derived from a conversation or question; company research may send vacancy text. Auto and fallback may contact multiple configured providers. Local models process their own inputs locally, not every feature in the app. Read the Privacy Policy.
Privacy controls stay in your handsScreen sharing and optional network settings
- Talkaside asks Windows to exclude the suggestion panel from recording and screen sharing. That is not a guarantee for every capture tool; check your sharing preview before a call.
- DNS-over-HTTPS and your own Shadowsocks, SOCKS5 or HTTP proxy are optional. These controls apply to Talkaside and leave the rest of Windows unchanged.
No hidden collectionNo analytics; a limited licence check
- Talkaside does not automatically upload call content, history or diagnostic logs to the developer for analytics or training. Online providers have their own data policies.
- Licensing uses licence, installation or pseudonymous device identifiers and licence dates, not call content. Support and payment records are separate. Every release includes a list of components for review.
- Downloaded models are optional. The Windows uninstaller keeps personal data and models by default, with an option to remove Talkaside-managed models; shared caches and personal files stay.
Your Windows calls. Your choice of AI.
Use one app for an interview or a work meeting in another language, with the AI providers and data flow under your control.
Talkaside captures audio from Windows calling apps without joining the meeting. History is stored on your PC. You choose the speech, translation and AI services; connected cloud providers receive the data needed for their part of the work and bill you separately.
Talkaside 1.0.4 supports cloud providers and local models: Whisper or Parakeet for speech recognition, Hy-MT2 for translation, and RapidOCR for screen text. AI answers can use cloud services or an existing Ollama or LM Studio server. Install optional local components from settings; available languages depend on each engine.
What people ask before installing
Will the other person hear or see that I am using it?
The app sends nothing into the call on its own. The only thing that goes into the call is what you turn on — the translated speech, if you enable it. The suggestion panel is an ordinary window on your computer, and you decide whether it appears when you share your screen. Test your own setup in advance, and if your company's rules forbid outside assistance, don't use it.
What else do I need besides the app?
Windows 10 or 11, 64-bit, plus access to your chosen speech and AI services. Cloud services need your own keys or a supported login. Local options need compatible hardware, disk space and downloaded models; local AI uses a server you already run, such as Ollama or LM Studio. Available options depend on your installed version. VB-CABLE is only needed to send speech back into the call.
Does it work with my calling app?
Talkaside captures microphone and call audio through Windows, without a meeting bot or a plugin for each calling app. Use it alongside Zoom, Google Meet, Microsoft Teams, Discord or browser calls. Check your app and audio devices with the five-day demo before an important call.
Where do my recordings and transcripts go?
Stored calls stay on your computer, not on a Talkaside call server. Cloud speech providers receive audio; online translation and AI receive relevant text and context, including material used for summaries or reviews. Local models process their own inputs on your PC. Provider policies apply to any data you send them.
How fast does a suggestion appear?
Talkaside shows recognised speech and streams answers as they arrive. The delay depends on the speech service, AI model, connection and length of the question. Local processing also depends on your hardware. Test your chosen setup before relying on it in a call.
Which providers can I use?
Speech recognition supports Deepgram, Google Gemini, Speechmatics, and local Whisper or Parakeet. Translation supports DeepL, Google Gemini, Groq and local Hy-MT2. AI answers can use supported CLI tools, cloud APIs or an existing Ollama or LM Studio server. Each recognition, translation and voice engine supports its own set of languages. Search by native, English or Russian names, or by language code. Use Test in settings to check your setup.
Can I use a VPN or change DNS without affecting Windows?
Yes. Advanced settings can send Talkaside's DNS lookups through DNS-over-HTTPS, or route the app and supported CLI providers through your own Shadowsocks, SOCKS5 or HTTP proxy. Both are optional and off by default. They do not change the rest of your computer's connection.
Can it see what is on my screen?
The capture shortcut takes a snapshot for a question. Select a monitor, window or region and choose local RapidOCR text recognition or image analysis. Online AI may receive extracted text or the image, depending on that choice. Screen recording is a separate option: when enabled, it continues recording until stopped.
Can it help in a job interview?
Interview mode drafts short and detailed answers using the conversation, your CV, the job description and any questions and answers you import. Check the suggestions against your actual experience; AI can get facts wrong. Use live assistance only where the interviewer's rules allow it. You can also use the app for preparation, translation and reviewing a recorded interview.
What happens if my connection drops mid-call?
Cloud speech recognition, translation, AI and voices need internet access. Switching cloud providers cannot fix a complete outage. Installed local components can handle their own tasks offline; local AI also needs its server running. Translation Auto tries DeepL, Google Gemini, Groq and installed Hy-MT2 in that order, skipping unavailable options. This may add a delay. Activation and licence requirements are separate from provider availability.
How do I buy a licence?
Start with the five-day demo. Use Buy in the header to open the Telegram bot, where you can see current plans, prices and payment methods. After payment, send your Machine ID from the activation screen; your licence key arrives in the same chat. Your AI, cloud speech recognition and paid voice services are billed separately by those providers.
Can I install it on a second computer?
A paid licence covers one computer. If you replace it or your hardware fails, contact us through the Telegram bot to have the licence reissued. For simultaneous use on another computer, arrange a separate licence through Buy. Each computer can try its own five-day demo; reinstalling Windows does not reset that period.
Why does Windows warn about the installer?
SmartScreen checks download reputation and can warn about an unfamiliar installer. A warning alone does not prove that a file is safe or malicious. Download from this site, check the file and scan it before deciding whether to run it. Warnings may appear again with another download or release; their behaviour depends on Windows and your security settings.