Skip to main content
Profy supports voice input, allowing you to communicate with the Expert by speaking. Your speech is transcribed to text in real time and then sent to the Expert as a text message.

​
How to Use

1

Click the Microphone Icon

Find the microphone icon on the right side of the input box and click to start recording. On first use, your browser will request microphone permission — click “Allow” to proceed.
2

Start Speaking

Once you see the recording indicator, speak your request into the microphone. Your speech will be transcribed to text in real time, and you can see the transcription in the input box.
3

Finish Recording

When you’re done speaking, click the stop button to end the recording. The transcribed text will automatically appear in the input box.
4

Edit and Send

After the transcribed text appears in the input box, you can edit and correct it before sending. Once confirmed, press the send button.

​
Supported Languages

Voice input supports multiple languages, including:
  • Chinese (Mandarin)
  • English
  • Mixed Chinese-English
If your speech includes mixed Chinese-English technical terms (e.g., “Help me optimize the performance of this React component”), the voice transcription can automatically recognize and correctly handle both languages.

​
Voice Input vs Text-to-Speech

Profy’s audio capabilities are divided into two independent directions: Voice input is how you express your needs to the Expert — the Expert’s replies are still in text by default. If you want the Expert’s reply to also be played as audio, click the read-aloud icon next to each reply.

​
Usage Tips

The accuracy of voice transcription depends on the clarity of your pronunciation. It’s recommended to use it in a quiet environment to avoid background noise affecting recognition.
For longer requests, consider breaking them into several segments. Express one complete idea per segment for better transcription results.
After voice transcription, you can manually correct any recognition errors in the input box, or add content that’s difficult to dictate (such as code snippets, links, etc.). Voice and text input can be freely mixed.
Recognition of technical terms and abbreviations (such as API, CSS, JSON) may be inconsistent. If you notice inaccurate transcription, simply correct it manually before sending.

​
FAQ

Please check whether your browser has granted microphone permission. There is usually a permission management icon on the left side of the browser’s address bar — make sure Profy’s microphone permission is set to “Allow.”
The platform enforces no duration cap — recording continues until you stop it manually. The seconds counter in the UI is a display only; it never truncates the recording.That does not mean you can record indefinitely, though. Transcription uploads the entire audio clip in one request, so the practical ceiling is the upstream transcription service’s per-request size limit (commonly around 25 MB). Exceeding it causes the transcription to fail rather than the recording to be cut short, so split long content into segments.
No. The transcription endpoint has no billing hook — it deducts nothing.Credits are consumed only afterwards: the transcribed text enters the conversation, and that turn is billed at your selected model’s token rate. So voice input costs exactly the same as typing the same words.

​
Dictating well: a worked example

Transcription handles continuous speech far better than a hesitant list. Compare:
The rule is one complete idea per take. If you need to think mid-sentence, stop the recording, think, and dictate again — the input box keeps what you already dictated, so you can append rather than restart. For anything with exact characters — code, URLs, file paths, API names — type it. Dictating https://api.example.com/v1/users?id=42 will cost you more time in corrections than typing it would have.

​
Failures and fixes

Voice input requires an active network connection. Voice transcription is not available in offline environments.

Speech synthesis and recognition

The billed ASR/TTS tools, which are separate from free chat dictation

Credits

How the resulting conversation turn is charged
Verified 2026-08-11. Sources: apps/web/src/hooks/useAudioRecorder.ts (no auto-stop timer), services/core/src/routes/audio/index.ts (POST /transcribe has neither a size check nor a billing call).