PixelOkay
Speech Studio
Browser-firstAll tools
Tools
✦ THE CREATOR WORKSPACE

Speech Studio

Capture speech, transcribe recordings, refine subtitles, and make text speak — in one calm, connected workspace.

01 · Capture02 · Transcribe03 · Caption04 · Speak
◉
Browser featuresLive mic, subtitles, voices · support varies
↓
Downloaded AI modelAudio transcription · local inference, internet for first load
⌁
Transparent privacyBrowser speech recognition may use a remote service
Check which speech and model APIs are available on your device.
WORKSPACE 01

Live speech recognition

Turn your microphone into a writing canvas.

BROWSER API
Recognition engine depends on your browser. Standard recognition may send microphone audio to a browser vendor's online service. Select “On-device only” to require local processing; that option needs browser support and may download a language pack.
On-device language availability has not been checked.
◉
Ready to listen00:00 elapsed
0 words

Live results have no reliable word-level timestamps.
Built for clarity, not mystery.Native feature availability is checked at runtime. AI model download is opt-in. No speech files are uploaded to PixelOkay servers by this implementation.

Using Speech Studio

Choose live recognition, audio-file transcription, subtitle editing or text-to-speech playback. The optional transcription model is downloaded only when you choose that mode.

Browser recognition and system voices may use remote services. Downloaded-model inference and GPU support depend on your browser and device. Native text-to-speech playback does not itself provide a downloadable audio file.

Keep spoken playback, transcript text and timed captions distinct

Practice script: The draft is ready for review.
Please confirm the owner before publishing.
  1. Use this authored script in Text to speech and choose an available system voice. Speak text is a playback action; it does not provide a downloadable narration file.
  2. For a separate listening exercise, open Live speech, select the language and recognition mode, then read the script only after granting the requested microphone permission. Compare recognized words with the original and edit errors manually.
  3. If using your own audio file instead, choose Transcribe audio and the intended model/device. The optional Whisper model download is separate from browser-default live recognition.
  4. Review transcript wording before opening subtitles. Inspect cue timing against attached media; editing words does not establish that all existing timestamps remain correct.
  5. Save project to preserve supported text, cues and preferences. Keep source media separately and reselect it when reopening; use TXT for reviewed wording and SRT/VTT for timed subtitle delivery.

Original script and planned exercise, not synthesized audio, recognized speech or model output. Browser-default recognition and some system voices may use remote services; on-device recognition requires actual browser support and language availability. No recognition or model inference was run for this guide. Use Subtitle & Caption Studio for its separate multi-track workflow, or Audio Editor Studio for editing an independently available audio file.

Choose dictation, transcription, subtitles or spoken playback

Speech Studio questions

Are transcript-to-subtitle timings automatically synchronized?

Not for the plain-text draft action. It divides text into caption-sized chunks with placeholder timings. Review every start/end time against the source recording. Where transcription provides timed segments, those still need accuracy and synchronization checks.

What should I save when continuing later?

Save project JSON for supported text, cues and preferences, TXT for transcript wording, and SRT or VTT for timed captions. Keep the source recording separately and reselect it when needed. A text export does not contain the original audio or a generated voice recording.

Related tools