If a draft isn't working — wrong tone, missed a topic, too long, too short — regenerate it. As often as you want, until you finalize.
Steps
Open the session in the note editor (it must be in Transcribed or Notes Ready status, not Finalized)
Click Regenerate note at the top of the editor
The current draft is replaced as the new note streams in
The transcript stays untouched. Only the generated note is replaced.
When to regenerate vs edit
Edit for small fixes: a word, a sentence, a section
Regenerate when the whole shape is wrong, or when you've changed a setting that affects the AI
Adjusting settings before regenerating
Common changes that warrant a regen:
Note format — switching from SOAP to DAP or back. Set in Settings → Notes → Note Format, or per patient on the patient detail page.
Detail level — brief, standard, detailed. Set in Settings → Notes → Note Detail Level.
Direct quotes — how often the note quotes the client verbatim. Set in Settings → Notes → Direct Quotes.
Modality — if the AI's clinical voice is wrong for this patient (you do CBT but the note reads like ACT), check the modality. The default is in Settings → Notes → Therapy Modality, with a per-patient override on the patient detail page.
Custom instructions — free-form text added to every note generation. Set globally in Settings → Notes → Note Custom Instructions, or per patient.
Change the setting, then regenerate.
Revising with a specific instruction
If you want to keep most of the existing note but change one thing — "make the assessment shorter," "add a safety statement to the objective" — you can give the AI a revision instruction.
In the regenerate flow, type your instruction in the prompt input that appears, then run. The AI receives the existing note plus your instruction and produces a revision rather than a fresh note from scratch. This preserves the parts you didn't ask to change.
Will regeneration take the same time as the first generation?
Yes — generation runs locally, no usage cost. The only "cost" is your time waiting for it. Wait times depend on your model tier and transcript length:
Session length | Pro on Apple Silicon | Lite on Apple Silicon |
Under 30 min | 30–90 sec | 20–60 sec |
30–60 min | 1–3 min | 1–2 min |
60–90 min | 2–4 min | 1–3 min |
90+ min | 3–6 min | 2–4 min |
Intel Macs and Windows PCs run roughly 1.5–3× slower than the figures above. The first generation after launching Confidant is also slower because the model has to load into memory; subsequent regenerations skip that step. Very long sessions (over 90 minutes) take an extra moment up front while Confidant processes the transcript in pieces — the progress label tells you which phase you're in.
See Generating a note from a transcript for the full breakdown.
What's preserved across regenerations
The transcript
Your patient settings (modality, format, custom instructions, case overview)
The session metadata
The diagnosis selection on the note (if you've already picked one)
The note text itself is replaced. If you've spent time editing a draft and want to keep parts of it before regenerating, copy the parts you want to a scratch buffer first — there's no merge.