What the Voice Workshop is
The Voice Workshop is Riterly's onboarding tool for building a writing profile. It analyses real writing samples you provide — across seven defined dimensions — and extracts the specific patterns that characterise how you write. The result is a writing profile that reflects your actual voice, not a description of it.
Why samples beat self-description
Most voice setup tools ask some version of the same question: how would you describe your writing style?
The problem is that the honest answer is almost always useless. Writers tend to reach for the same words — "friendly and approachable," "conversational but professional," "clear and direct." These phrases describe most bloggers. They give the AI almost nothing to distinguish one writer from another.
This isn't a failure of self-awareness. The patterns that define how someone writes — sentence rhythm, structural habits, the specific way they open a paragraph or close an argument — are largely unconscious. Writers don't notice them because they're not decisions anymore. They're defaults.
Real writing samples reveal what self-description cannot. The way you actually use subordinate clauses, whether your vocabulary shifts between technical and plain registers, how long your paragraphs run before you break them — none of that shows up in a questionnaire. It shows up in the writing itself. That's what the Voice Workshop reads.
The four phases
The workshop takes about two to three minutes of your time and runs through four phases.
Phase 1: Niche Selection
The first step is selecting a niche. This loads a niche template — a pre-built profile for your content type that acts as the structural foundation for your writing profile. The niche template defines the baseline; your samples refine it.
You also set a profile name and pick a colour for easy identification on the dashboard. These are small decisions, but getting the niche right matters — it shapes the structural layer the AI builds your patterns on top of.
Phase 2: Upload Writing Samples
This is where you give the system something to read. Riterly accepts .txt, .md, .doc, .docx, PDF, and URLs. For URLs, Riterly fetches the content automatically — you don't need to copy and paste.
Once uploaded, each sample appears as a chip in the interface. You can remove any chip before proceeding if you change your mind about a sample.
Before moving to synthesis, there is a one-time acknowledgement step confirming that the content belongs to you or that you have permission to use it.
The minimum useful sample is one post of a few hundred words — enough to give the AI something to work with. The optimal range is five to ten posts spanning different topics and styles. More variation in the samples produces a more accurate profile.
Phase 3: AI Synthesis
Once you submit your samples, the synthesis runs asynchronously. You don't need to wait on screen — you can navigate away and come back. This typically takes one to two minutes.
During synthesis, Riterly extracts seven dimensions from your writing:
- Tone — the overall register and emotional quality of your writing
- Sentence rhythm — how your sentences are structured and how they vary in length and complexity
- Vocabulary level — the range and specificity of the language you use
- Structural patterns — how you organise and sequence ideas
- Narrative style — how you tell stories or build arguments
- Engagement style — how you address and connect with the reader
- Distinctive patterns — specific habits or tendencies that stand out as yours
Phase 4: Review
When synthesis is complete, Phase 4 shows each dimension with its extracted value and a confidence indicator — High, Medium, or Low.
Low confidence means the AI found mixed signals in your samples or was uncertain about a pattern. These are the dimensions to check most carefully. The confidence indicator is not a quality score — it's a signal about where the analysis needs your attention.
Phase 4 uses a rich text editor, so you can rewrite any section, remove findings that don't ring true, or add things the AI missed. Once you're satisfied, clicking Confirm creates the profile.
What to use as samples — and what to avoid
Good samples are posts you wrote yourself, across a range of topics and styles. Variation matters — the more the samples reflect the breadth of how you actually write, the more accurate the profile will be. Five to ten posts is the target. One is enough to start.
Avoid AI-assisted writing. If you submit content that was generated or substantially rewritten by AI, the synthesis will model the AI's patterns, not yours. The profile will reflect the tool, not the writer.
Avoid collaborative pieces where your voice has been diluted by an editor or co-author. If the final text doesn't sound like you writing alone, it won't produce an accurate read of your patterns.
After the workshop: the Profile Sandbox
Confirming the profile in Phase 4 is not the finish line. The profile the workshop produces is a starting point.
After you confirm, Riterly takes you directly to the Profile Sandbox — and that's where the actual calibration happens. The sandbox lets you generate test drafts and compare them against your expectations. Two to four runs is the normal range before a profile reaches a state that feels right. That's not a sign anything went wrong — it's how the system is designed to be used.
A few other things worth knowing:
- Your samples are not retained. They're stored temporarily for the duration of the synthesis job, then deleted. They're not used for any other purpose.
- Rebuilding the profile means starting a new profile through the workshop wizard. Your old profile's version history is preserved — nothing is deleted. You're adding, not replacing.
- Direct editing is available through the advanced markdown editor, which gives you access to the full profile text if you want to make changes outside the workshop interface.
After you confirm your profile, open the sandbox. Generate a draft. Read it against something you've actually written. That comparison is where you'll learn what the profile got right, what it missed, and what to adjust next.