Privacy Policy

Cogumi Voice is a set of browser-based voice tools. There is no account system, no payment system, and no server-side database for user notes, conversations, uploaded files, or generated songs. Some features still need network requests to work; this policy explains those cases clearly.

Summary

• Dictate, AI Spirits, and optional local AI features run model inference in your browser. Their notes, chats, and generated content are stored locally in your browser unless you choose to submit something through the contact form.
• Web Reader, Document Reader playback, and Dictate playback may send short text chunks to the Cogumi Voice Cloudflare Worker so speech audio can be generated.
• Google Analytics loads only after you select Accept in the analytics banner. If you select Decline, Analytics is not loaded for that browser.
• The contact page embeds Google Forms. Anything you submit there is handled by Google Forms and delivered to the site operator.

Information You Provide

Depending on the tool you use, you may provide article URLs, document files, pasted text, dictated speech, notes, prompts, chat messages, song text, or contact form responses. Cogumi Voice uses this information only to provide the tool you requested.

What Stays In Your Browser

• Dictate: speech cleanup, summarization, notes, and summaries are stored locally in IndexedDB. Model inference runs locally through WebGPU or WASM.
• Dictate speech recognition: desktop browsers may use their built-in Web Speech recognition service, which may process microphone audio through the browser vendor. Mobile dictation uses the downloaded multilingual Whisper Tiny ONNX model directly, and desktop Dictate switches to that local model if built-in recognition is unavailable or fails. Whisper audio processing runs locally in your browser through WebGPU.
• AI Spirits: companion profiles and conversations are stored locally in IndexedDB. Chat and local voice features use downloaded browser-side models where available.
• TweetTune: OddVoices singing, instant backing synthesis, optional MusicGen backing generation, and song history run or remain in your browser. Model files may be downloaded through the same-origin proxy described below.
• Document Reader extraction: uploaded files are parsed in your browser. The full file is not uploaded to a Cogumi Voice document-processing server.
• Model files: local AI model files may be downloaded through the same-origin /hf/ proxy and cached by your browser Cache Storage for later visits.
• Preferences and temporary state: browser storage such as IndexedDB, Cache Storage, local/session storage, and the consent cookie may be used to keep local app state.

What May Leave Your Browser

• Text-to-speech: when you press play for Edge voice output, Cogumi Voice sends bounded text chunks, voice selection, and speed/rate settings to the Cloudflare Worker endpoint /api/edge-tts. The Worker requests speech audio from the upstream speech service and returns audio to your browser.
• Web Reader URL fetches: when you paste a URL, the Worker endpoint /proxy fetches that page so the browser can extract readable text despite CORS restrictions.
• Contact form: the contact page is a Google Forms embed. Form submissions can include whatever name, email address, message, or other details you enter.
• Avatar generation: AI Spirits may request DiceBear avatar SVGs using a style and seed. Do not put sensitive information in avatar names or seeds.
• Analytics: if you accept Analytics, Google Analytics receives usage data such as page views, device/browser information, approximate location derived from IP, referrer, and interaction events configured for this site.

Cookies And Analytics

Cogumi Voice stores your analytics choice in a first-party cookie named cogumi_cookie_consent. The value is either accepted or declined and is kept for up to one year unless you clear browser cookies.

Google Analytics is not loaded until you select Accept. Google requires sites using Google Analytics to disclose Analytics use and how data is collected and processed. Google’s own policies explain how Google may process Analytics data.

Service Providers

Cogumi Voice relies on third-party providers to operate the site and selected features:

• Cloudflare: hosting, Worker execution, asset delivery, request routing, and security logs.
• Google: Google Analytics after consent and Google Forms for contact submissions.
• Microsoft Edge text-to-speech service: speech synthesis for Edge voice playback requested through the Worker.
• Hugging Face/model hosting: source for model artifacts that are proxied through Cogumi Voice where configured.
• DiceBear: generated avatar images for AI Spirits.

Retention

• Local notes, conversations, saved songs, model caches, and preferences remain in your browser until you delete them through the app or clear browser/site data.
• Cogumi Voice does not maintain a user account database or a cloud copy of your local notes or chats.
• Contact form responses are stored in Google Forms according to the Google account/form settings used by the site operator.
• Cloudflare and other service providers may retain operational logs according to their own policies and service settings.

Your Choices

• Select Decline in the analytics banner to keep Google Analytics off.
• Clear the cogumi_cookie_consent cookie if you want to be asked again.
• Use browser settings to clear local site data, including IndexedDB, Cache Storage, local storage, and cookies.
• Do not submit sensitive information into URLs, contact forms, prompts, notes, chats, or text-to-speech content unless you are comfortable with the feature’s data flow described above.

Children

Cogumi Voice is a general-purpose tool and is not directed to children. Do not use the contact form to submit personal information about children.

Changes

This policy may be updated when Cogumi Voice adds features or changes providers. The effective date will be updated when material changes are made.

Contact

Questions about this policy can be sent through the contact form.