Shape an original synthetic voice from words, or prepare two recordings of your own voice. Free-first, private-by-default, portable Markdown. No API key, account, or upload required for this prototype. Download 24 kHz mono WAV files for Qwen3-TTS.
Local draft, not a synthetic voice generator. This page can prepare descriptions and microphone recordings. Actual OpenAI voice creation happens separately through an approved account/API. Do not impersonate celebrities, actors, or other people, and do not upload their archival recordings.
1. Design an original voice
Initial preset: The Curious Storyteller, inspired by general qualities of expressive, warm conversational storytelling. No identity or character likeness is requested.
0 / 2000 charactersCopy into the OpenAI prompt-to-voice workflow
Copy the description, open the official Voice Playground, and paste it into the voice-description field. This app does not transfer text to OpenAI automatically. Voice creation is free for API customers; hearing it through GPT-Live is billed separately from ChatGPT Plus. Official instructions ↗
Creating a prompt-generated voice currently works for GPT-Live. OpenAI creates a saved voice ID without returning preview audio. Listen in a billed GPT-Live session or the official voice Playground. A browser's built-in speech reader is not a preview of the described synthetic voice.
2. Record your own voice
Read the on-screen text, use your microphone, listen locally, then download a 24 kHz, mono, 16-bit PCM WAV file ready to upload as a reference sample in Qwen3-TTS. Nothing leaves this browser page automatically. Conversion happens on your device after you stop recording.
Checking microphone availability…
You can record in a normal browser tab that allows microphone use. Some embedded file viewers prohibit it.
To record outside this viewer: Download and unzip the complete Voice Studio package. Open aletheia-voice-studio.htm directly in Chrome or Edge (not a preview pane), then select Allow when prompted. If local-file access is blocked by your browser, host it on HTTPS (such as GitHub Pages) or localhost. Audio remains in your browser until you choose Download.
This must be a separate recording in your own voice. Check OpenAI's current approved wording before submitting it.
I am the owner of this voice and I consent to OpenAI using this voice to create a synthetic voice model.
00:00
B. The Raven · Edgar Allan Poe (1845)
First stanza, public domain. Read in your own natural voice with pauses and expression. About 40–60 seconds depending on pace. This is not a command to impersonate any actor or poet. Read the full poem ↗
Once upon a midnight dreary, while I pondered, weak and weary,
Over many a quaint and curious volume of forgotten lore—
While I nodded, nearly napping, suddenly there came a tapping,
As of some one gently rapping, rapping at my chamber door.
“’Tis some visitor,” I muttered, “tapping at my chamber door—
Only this and nothing more.”
Read at a comfortable pace for 10–60 seconds; the recorder stops at 60 seconds. For a faithful standard speech assessment, use the linked original rather than this alternative.
00:00
Personal voice recordings are sensitive biometric material. Keep both files private. Do not commit them to a public repository. Recording them is not equivalent to OpenAI accepting or creating your voice. Recording-based voice creation currently requires approved API access.
3. Workflow, limits & evidence
From description
Edit an original adult voice description of at most 2,000 characters.
Copy the description into an authorised prompt-to-voice API client or Playground.
Use the saved voice ID with GPT-Live. Creation free; listening sessions billed.
Keep the description in portable Markdown.
From your own recording
Record exact consent and choose either The Raven or the rainbow reading as a 10–60-second sample. The two passages can also be recorded and downloaded separately.
Download and listen locally to both recordings as 24 kHz, mono, PCM WAV, which is compatible with Qwen3-TTS reference-audio upload.
For Qwen3-TTS, upload the WAV in Voice Clone (Base) ↗, with the corresponding spoken text in the Reference Text field (unless x-vector-only mode is selected). OpenAI requires separately approved access for recording-based voice creation.
Submit only through the permitted API workflow on a trusted server.
Aletheia Storyteller ↗ already handles reading stories aloud. This proposed Voice Studio does not replace it.
This local build has not been published or tested across actual microphones/browsers. Chrome, Firefox and Safari can differ in recording permissions and supported file formats. No playback or synthesis API has been called.