Sudowrite for academic fiction - does it handle period dialogue?

I'm evaluating AI writing tools for a new project: a historical fiction novel set in the 17th century. My primary concern is dialogue authenticity. Most general-purpose LLMs fail at period-appropriate speech, either defaulting to modern slang or a generic, faux-Shakespearean mess that reads like a bad fantasy novel.

I've run Sudowrite through a series of cost-per-quality tests, treating it like a cloud service. My initial experiment used the "Describe" and "Brainstorm" features on a simple prompt: "A merchant and a ship's captain arguing over the price of nutmeg in a London tavern, 1652." The output was serviceable but leaned heavily on cliché. The real test was using "Rewrite" and "Expand" on dialogue I had manually crafted.

Here are the raw results from a dialogue snippet test:

**Original Snippet (My Baseline):** "The price is a scandal, sir. Four shillings the pound? I had it from the factor at three and six last fortnight."
**Sudowrite 'Rewrite' (Tone: 'Authentic, 17th century'):** "Good sir, the sum you name is a grievous imposition. Only in the last month did I pay but three shillings and sixpence per pound to your own factor."

The tool changed the structure but introduced "grievous imposition," which feels anachronistically formal. It also made the time reference vaguer ("last month" vs. "last fortnight"), which loses specificity.

My workflow assessment:
* **Strengths:** The "Visualize" feature is unexpectedly useful for period objects and settings, reducing my research time. The "Write" function provides a decent first draft for descriptive passages.
* **Weaknesses:** Period dialogue requires heavy manual correction. The AI often overcorrects, making speech stilted. It has difficulty with class-based dialect variations (e.g., merchant vs. sailor).
* **Cost Analysis:** At approximately $0.02 per "Rewrite" operation on a paragraph, the iterative cost of refining dialogue adds up. It's less efficient than using a well-tuned local model, but the convenience factor is high.

Has anyone else used Sudowrite for similar period-specific work? I'm particularly interested in:
* Your prompt engineering strategies for locking in a specific era's diction.
* Whether you've found the "Tone" settings (like "Authentic") to be consistently useful or just marketing.
* Any comparisons to fine-tuned models or other services in terms of output quality versus time/currency investment.