Voice & knowledge
Profile-wide evidence for voice, facts, and boundaries
Voice & knowledge is the active Synfluencer's persistent context library. Add writing, email, documents, webpages, transcripts, recordings, or video once, then control whether the material may influence voice, supply reusable knowledge, or remain a private library item.
Why it is separate from From Source
- Voice Library is global to the profile. Approved self-authored evidence helps every applicable generation sound consistent. You never pin voice evidence to one draft.
- Knowledge can be automatic or draft-specific. Turn on Use when relevant for facts and guidance that may apply broadly, or select a library item from From Source's Knowledge tab when only one draft needs it.
- From Source is temporary composition context. Its pasted text, links, files, and selected Knowledge items stay attached to that draft in Review; they do not silently change the profile.
Add approved voice evidence
- Open Voice & knowledge below the profile selector in the sidebar.
- Choose Voice Library, then paste text, add a public URL, or upload a supported file.
- Identify authorship and confirm you have permission to use the material.
- For mixed-speaker recordings, confirm the profile speaker or select only the relevant transcript passages.
- Approve the source. Third-party material cannot become voice evidence.
- Select any pasted writing item later to review and edit its full title and content.
Maintain reusable knowledge
Use Knowledge for product facts, positioning, terminology, customer context, approved claims, legal boundaries, and other material that should keep future drafts accurate. Review old items, disable automatic reuse when guidance expires, and delete material that should no longer be available.
Navigate a large library
The library shows six items at a time on desktop and mobile. Search by name or filter by file type and upload date; the visible range and page controls always reflect every matching stored source.
Privacy and retrieval
The library can grow without a five-item cap, but each generation receives only a small, relevant subset. Source bodies are editor-only. Audio and video use the configured zero-data-retention transcription routes, and imported library material is not silently sent to an embedding provider.