Corti’s text generation capabilities turns raw clinical inputs — conversations, dictations, documents — into reliable, structured clinical notes. Two core capabilities work together: facts and synthesis. Everything else is designed with the flexibility to build custom-tailored and scaleable workflows to achieve clinical-grade quality documentation.
Start with facts (optional, but valuable)
As a conversation unfolds, Corti’s FactsR can extract key clinical facts in real time — symptoms, findings, medications, allergies, and so on. Facts are optional: you can generate a note directly from the transcript without them. But when they’re present, they give the downstream synthesis a cleaner, validated foundation to work from. Fewer missed details, increased conciseness, more complete notes.
→ FactsR overview
Synthesize the note
Once you’re ready to produce a note, Corti’s synthesis endpoint accepts several kinds of input: extracted facts, a raw transcript, prior documents — or any combination. The endpoint reads what you give it and produces a note shaped by the structure you’ve defined.
→ Guided Document Synthesis
Sections and templates
This structure comes from sections — small, reusable pieces like History of Present Illness, Assessment, Plan, Vital Signs. Each section has its own prompts telling the model what content to include, in what voice, and in what shape.
Sections can be grouped into templates (a SOAP note, a referral letter, a discharge summary), but they don’t have to be — you can assemble them ad-hoc for a single call. The overall structure is governed by a mix of programmatic elements — schemas that fix the shape of the output — and prompts that live inside each section.
→ Sections & Templates overview
Sometimes free-text prose is what you want. Sometimes the note’s output needs to feed a structured data pipeline — an EHR field, an analytics table, a coding step. Sections carry a typed output schema modelled after standard JSON that can constrain a section’s output to a string, number, boolean, array, or object. Downstream systems get parseable JSON, not a free-text blob to regex.
What sets Corti apart, is that the same resources and endpoints give you both the string version and structured version of the generated document. Plus, we’ve built in extra programmatic control of the string version to ensure robust fall-backs, placeholder texts and similar use cases where prompting can be fragile or cumbersome
Standards, customize, or author your own
To make it easy to get going, Corti maintains a growing library of Standards — sections and templates covering the most common clinical use cases across many specialties and languages. You can:
- Use a Standard as-is — reference it by ID and go.
- Tweak a Standard — inherit from it and override only the fields you want to change (heading, prompts, or output shape).
- Author your own from scratch — full control over every prompt and every field.
Every section and template is a versioned, publishable resource with a proper lifecycle. You can iterate safely; the previously published version stays live until you promote a new one. Build template management into your application by leveraging filters on structured labels, languages, regions, and specialties.
→ Corti Standards library
Provide your customers the right note structures
When you build with Corti’s embedded Assistant UI, then you can manage templates in the Console and assign them easily to organizations such as your customers or different departments for end-user availability of sections and templates.
If you don’t, you can control access on your client application side but still leverage the Console Template Builder & Management: any section and template resources created via API by one of your Direct API Clients can be seen and managed in Console.
→ Manage Templates in Console
Enable your customers or even end-users to customize
You can leverage the same setup we use to author and maintain the Corti Standards, to author your own standards to offer customers, and then more.
Let customers tweak your standards (or Corti’s) and store as their own resources for repeated use – showing them the full prompts to tweak or abstracting away with presets you define they can apply.
You can also handle template and access management entirely on your side and use one of the suitable docGen request shapes.
→ Customization Guide
Four ways to generate a document such as a clinical note
When you’re ready to generate the note, the same endpoint gives you flexibility depending on where you are in your journey or your workflow setup:
- Reference a stored template by ID — the lightest path, easy to try and fit for template-centric solutions.
- Reference a template with runtime overrides — patch a section or two for a single call without persisting a new resource.
- Assemble sections on the fly — build a template inline from existing section references – overrides are also here supported.
- Define everything inline — define all sections with their schema and all prompts in the request body, no persisted resources needed. Ideal for prototyping or when handle end-user templates entirely on your end.
→ The four template-supply paths
Where to go next
Corti Assistant