English privacy policy
Boundless Flow is operated by 齐码蓝智能科技（大理市）有限责任公司, under the ZimaBlueAI brand. Contact bf-support@zimablueai.com for privacy requests, support and AI-content reports.
Local processing and optional services
Local-model inference runs on your device. If you enable a configured model service or API, the audio, text, reference voice, prompts and parameters required for that task are sent to that endpoint, which may be local, on your network, or operated by a third party abroad. Its retention, training, fees and privacy terms apply separately. The provider can normally observe your request's network source address. Review the endpoint and provider before sending content; use HTTPS for public endpoints. Models and components are prepared separately; download providers process their own network requests.
Information used
Depending on the features you use, the application processes recordings and imported audio; transcripts, corrections, translations and project documents; reference voices, speaker features, labels and emotion estimates; settings, model paths, endpoints and API credentials; and task states, timestamps, file paths, errors and local records. These support the selected transcription, input, meeting, report, translation, synthesis and analysis tasks. Local speech inference itself does not require uploading that content to us. This policy does not authorize us to use your content for model training; review third-party providers' terms separately.
Recording, input and analysis
Recording tasks you start use the microphone subject to Windows permissions. Speech input may deliver text to another application; check the target before using it. Closing a panel or window may leave a recording, interpretation or announced background task running; use its Stop or Cancel controls. Obtain the appropriate rights and permissions before using other people's recordings or voices. Speaker and emotion results and AI summaries may be wrong; they do not authenticate identity or provide medical or psychological diagnoses.
Synthesis provenance
Some integrated synthesis, voice-design and cloning outputs automatically include a low-perceptibility provenance watermark. The current integration seals a generation identifier and retains related identifiers, timestamps and audio hashes locally. Platform-held offline tools and private keys can verify the corresponding payload. This workflow does not automatically upload the ledger or collect a public IP address, MAC address or hardware serial number; client software does not download platform private keys. Live speech-translation output is excluded from this watermark workflow. Original reference audio and model files are not modified by it. Identifiers can link records and outputs and are not necessarily anonymous. Detection does not authenticate a speaker or operator or establish the truth of the content. Absence is not proof of human speech; transcoding and editing may prevent detection.
Storage, retention and your controls
Data is stored, as appropriate, in application user-data folders and project, model, output or backup locations you choose. Some tasks retain source audio, derivatives or provenance records. Local data normally remains until you delete it or a relevant feature clears or replaces it; not all records have automatic expiry. Clearing a view, removing a list entry or uninstalling is not a guarantee that every recording, export, backup or cache is erased. Stop active tasks and check storage locations before cleanup. Use available application controls and operating-system tools to manage files, disable features or permissions, change services and remove credentials. Exports and recipients can retain separate copies. Contact third-party services about data they retain.
We cannot remotely access or erase data held only on your device. If you voluntarily send support material or a privacy request, we use it to handle that matter, retain it as needed for that purpose and applicable requirements, and handle appropriate access, correction and deletion requests. Send only necessary information, not unredacted credentials, private keys or unnecessary recordings.
Security, minors and changes
Protect device access, filesystem permissions, backup locations and service endpoints. Encrypted watermark payloads do not mean every recording, setting or report is encrypted. Minors should use recording, cloning, speaker analysis and third-party services with guardian guidance and the necessary permissions. We update this policy when processing changes and provide any additional notice or authorization required before relevant feature use. This policy does not replace permissions required for individual recordings, cloning or data transfers.
This page and contact
Hosting and network providers receive web request information such as IP address, browser details and time. The product domain currently uses GitHub Pages; see the GitHub privacy statement. The company's mirror is served by its website server. This policy page contains no analytics scripts, advertisements, external fonts or submission forms. Email providers process support messages and attachments you choose to send. For privacy or data requests and AI-content reports, email bf-support@zimablueai.com with the request type, software version and only necessary context. We handle requests after checking necessary information and applicable requirements.