Dictation for Legal Briefs:
Replace Dragon Legal on Windows

Legal drafting is text-heavy work

Attorneys spend a substantial fraction of every working day producing written text. Briefs and memoranda for the court, contracts and amendments, client letters, deposition summaries, witness preparation notes, internal research memos, time entries, billing narratives, and the long tail of email and instant messages that surround all of it. Even attorneys who are fast typists end the day with sore hands and wrists. Voice dictation, when it works, can cut drafting time meaningfully and reduce the physical load.

The legal dictation market has been dominated for years by Nuance Dragon Legal (now Dragon Legal Anywhere, with the Nuance acquisition by Microsoft folded in). The product offers integrated legal vocabularies, courtroom and case-management specific terminology packs, document workflow integrations, and per-attorney voice profile management. It is priced for institutional buyers: typical perpetual licenses start in the 500 USD plus range per user, with annual maintenance on top, and enterprise editions go higher. For a large firm with full IT support, the institutional package can make sense. For solo practitioners, small firms, and lawyers who would rather not navigate enterprise procurement and per-user voice training to add dictation to their workflow, the economics often do not fit.

That is the gap StarWhisper is positioned for. It is a free local-first Windows dictation app powered by OpenAI Whisper. It is not Dragon Legal, does not pretend to be, but for the bulk of everyday attorney drafting (narrative briefs, memos, correspondence, notes, time entries) it covers the workflow at a small fraction of the cost.

What Whisper accuracy looks like on legal content

OpenAI Whisper was trained on roughly 680,000 hours of multilingual and multi-task supervised audio drawn from a wide variety of sources. In independent benchmarks, the large Whisper models reach 95 to 98 percent word accuracy on clean English audio. On legal narrative drafting, the model handles the typical vocabulary of practice areas like litigation, contracts, real estate, employment, family, immigration, and IP reasonably well. Common terms (statute, regulation, jurisdiction, plaintiff, defendant, motion, brief, deposition, interrogatory, summary judgment, fiduciary, indemnify, severability) transcribe correctly the first time in most cases.

Where Whisper is less reliable is dense citation-heavy passages. Long strings of Bluebook citations spoken at speed, unusual reporter abbreviations, pinpoint citations rendered as compressed alphanumeric, and very recent statutory amendments often need correction. A typical workflow for that is to dictate the narrative argument and then type or paste the citation strings, or to use a citation manager (Westlaw Quick Check, Lexis Shepard's, Bluebook citation tools) to assemble the citation portion separately.

For long-form drafting (a statement of facts, an argument section, a factual background section, a client memo, a draft contract clause) where the work is paragraph-level prose, Whisper accuracy is high enough to be a real time saver. See the professional accuracy page for a broader discussion of where Whisper holds up and where it does not.

Tool compatibility across a typical law firm stack

Because StarWhisper types text as keyboard input at the OS layer, compatibility is broad. The following are confirmed compatible based on field testing and user reports:

"Works" here means StarWhisper can be activated by hotkey while the cursor is in a text field and the dictated text appears as if typed. Structured controls like dropdowns, checkboxes, and field-driven document assembly templates accept dictation only in their free-text portions. For broader role context, see voice to text for lawyers.

Attorney-client privilege and the privacy posture

Attorney-client privilege protects confidential communications between attorney and client. Modern legal practice routinely involves a long chain of third-party software (email, cloud storage, case management, research databases) that handle privileged content. Most of those have established models for how to fit within the standard of care, typically through vendor agreements and security commitments. Cloud-based dictation tools are part of that chain: when a vendor processes voice audio of privileged communications, the firm has to evaluate whether that processing fits within its security and privilege protection program.

StarWhisper's local-only architecture sidesteps that evaluation for the dictation portion. In Local Mode, the Whisper model runs on your local CPU or NVIDIA GPU. The audio is captured by your microphone, processed in memory on your machine, converted to text in the focused application, and the audio buffer is discarded. There is no transmission of audio or transcript to any vendor, no cloud retention, and no third-party processor of the privileged content. The dictation tool does not introduce a new third-party intermediary in the way a cloud-based dictation service would.

Cloud Mode is opt-in, disabled by default, and is not appropriate for privileged communications. For firms deploying StarWhisper for privileged work, keep Cloud Mode disabled. IT can lock it off at install time so attorneys cannot accidentally enable it. See the HIPAA dictation page for the analogous medical privacy analysis, which uses the same architectural argument.

A typical drafting workflow with StarWhisper

The basic loop for drafting a brief or memo:

• Open Word and create your document with the firm's template, page setup, and styles.
• Position your cursor in the first paragraph you want to draft.
• Hold the configured push-to-talk hotkey, dictate the paragraph, release.
• The transcribed text appears in Word as if you had typed it.
• Edit with the keyboard the way you would correct any draft.
• For citation strings, type them or pull them from your citation manager.
• Repeat through the document.

For client correspondence, the same loop works in Outlook compose windows, Gmail in the browser, and Slack messages. For time entries, dictate the narrative directly into your time entry field in Clio, MyCase, or your billing tool. For research memos, dictate into Westlaw or LexisNexis note fields, or into Word with research tabs open in the browser.

Most attorneys end up with a hybrid workflow: dictate the long-form narrative portions (where typing is the slow part), and type the structured portions (case captions, citations, signature blocks, footnotes). The two combined feel faster than either alone.

Where Dragon Legal still wins, honestly

Several scenarios where Dragon Legal Anywhere or another enterprise legal dictation product remains the better fit. Matching tool to workflow matters more than winning a marketing argument.

• Built-in Bluebook and legal vocabularies. Dragon Legal ships with integrated legal vocabulary packs and citation handling that StarWhisper does not replicate.
• Voice-command document assembly. Dragon's macro framework can insert standard clauses, navigate templates, and trigger document workflow steps by voice. Whisper-based dictation does not include that.
• Court-specific terminology. Specialty areas (admiralty, patent prosecution, immigration, certain state codes) have vocabularies where Dragon's specialty packs save real time.
• Per-attorney voice profile management. Firm-managed profiles, central deployment, SSO, analytics across attorneys. Standard enterprise tooling Dragon Legal provides.
• Court reporting and stenography. A different category entirely. StarWhisper is dictation, not court reporting. CART and steno-style applications are not what this tool is.

StarWhisper is positioned for the attorneys who do not have or want those wraparound services: solo practitioners, small firms, attorneys at large firms who want a personal dictation tool for their own work without firm IT involvement, and attorneys frustrated with Dragon Legal's licensing complexity. The free tier exists so you can validate the workflow on your machine before paying anything.

Setup for a firm or solo practitioner

For individual attorneys:

• Install from the download page or the Microsoft Store.
• Confirm Local Mode is enabled (default) in Settings, Transcription mode.
• Choose a push-to-talk hotkey that does not conflict with your firm's macros or Word shortcuts.
• Test on a non-privileged document first (a scratchpad), then on real work.

For a small firm rolling out across multiple attorneys, the same flow scales. There is no per-attorney voice training, no profile migration, and no per-user licensing dance. Pro licenses can be purchased per attorney. For firms where IT manages the workstations, the administrator configuration option can lock Cloud Mode off at install time. For more detail on the broader Windows dictation landscape including a comparison to the legacy alternatives, see StarWhisper vs Dragon.