Kensho
Kensho Technologies (S&P Global) · Coding
Kensho Scribe is S&P Global's speech-to-text service for financial audio. Earnings call transcription arrives formatted, and both batch and realtime APIs are available.
About Kensho
What Is Kensho Scribe
Kensho Scribe is the business audio transcription service from Kensho, the AI unit S&P Global bought in 2018. Its model was trained on hundreds of thousands of hours of business and financial audio, labeled by S&P Global's own transcript team. That training choice is the whole pitch. Earnings calls, investor presentations, and analyst briefings come with tickers, currency figures, and executive names that general-purpose speech-to-text services often mangle.
What comes out is more than a stream of words. Scribe adds punctuation, stays consistent with speaker turns, and strips the fillers and false starts that make raw audio hard to read. You get a document that a person can skim, not something that needs heavy cleanup first.
Two constraints matter before you sign up. Coverage leans on English business audio, so teams working in other languages should confirm support before building a pipeline. And this is an enterprise service with seat-by-seat provisioning, not a self-serve product you buy with a credit card. Pricing comes from Kensho's sales team.
Getting Started
- Request access through the Scribe site and pick the right path. The web interface suits one-off files, while the API suits bulk work and automated workflows.
- Get an access token. Kensho's docs walk through authentication and token generation for the batch and realtime endpoints.
- Upload your audio. On the Batch API, send the file directly or pass a URL that Scribe can fetch without an auth header.
- Choose an output format. Ask for JSON when you need structured timestamps and speaker labels, or request Word documents for handoff to editors.
- Collect the transcript. Poll the transcription ID until the job returns a 200, or set a callback URI so Scribe notifies you when it's done.
Product Information
A quick look at Kensho's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Equity research analysts
- Investor relations teams
- Transcription vendors and expert networks
Tasks
- Transcribing quarterly earnings calls
- Building a searchable audio archive
- Producing live captions
- Preparing interview or deposition audio for editors
Scenarios
- Getting a few files transcribed without writing code
- Feeding earnings season volume through a pipeline
- Handling messy real-world audio
Key features
Finance-Tuned Speech Recognition
The acoustic model and language model were trained on S&P Global's own labeled finance audio, roughly 300,000 hours in the current release. That means earnings call vocabulary, ticker symbols, and executive titles get recognized rather than guessed. Models built for general conversation tend to trip on exactly these terms.
Automatic Formatting and Disfluency Removal
Scribe punctuates the text and removes filler words, false starts, and repetitions as it writes. The output reads like a transcript someone cleaned up, so downstream readers and natural language processing jobs don't inherit audio noise. For a research desk, that's the difference between a usable document and a rough dump.
Speaker Diarization and Identification
Separate models track when the speaker changes and work out who is talking. Analysts reading a call transcript can tell the CFO's answer from the operator's instructions at a glance. Diarization also matters when several people speak over each other during Q&A.
Batch API for Files and Remote URLs
Send audio or video with a POST request, either as an uploaded file or as a link Scribe fetches on its own. Each request returns a transcription ID, and the job runs in the background. Poll the ID or set a callback URI to learn when it's done, then pull the result as JSON or Word.
Realtime WebSocket Transcription
The realtime endpoint accepts uncompressed audio as it's produced and returns text during the session. Connections can run for up to five hours at twice real time upload speed, which covers most meetings and long webinars. It's the option for live captioning and in-meeting notes.
Human-in-the-Loop Escalation
When machine output isn't enough, the human-in-the-loop service adds professional transcribers who correct and format the result. You set a priority level tied to your contract, and turnaround windows for high, medium, and low priorities are agreed in advance. Output arrives as a Word document.
Pros and cons
Pros
- Accuracy on financial audio is the product's reason to exist, and it shows on tickers, company names, and earnings phrasing that general models mishear.
- Transcripts arrive formatted with speaker labels and timestamps, saving an editing pass on every file.
- Speed is strong: the current model processes one hour of audio in about ten seconds.
- Two access paths cover most workflows, with a web interface for one-off files and APIs for automation.
- The human-in-the-loop option gives you a fallback when a recording is too noisy for machine output.
Cons
- There's no public price list. You need a quote from Kensho's team, which slows down anyone comparing options on their own.
- Batch transcripts are only retained for up to two days, so you must pull results promptly or store them yourself.
- Throughput has ceilings you have to design around: 1 GB per batch file, five API requests per second, and a fixed audio allowance per account.
- The offering is built around English business and financial audio, so teams needing other languages should verify support first.
Frequently asked questions
Kensho Scribe is an automated transcription API and web tool that turns business and financial audio into formatted text. Earnings calls are the flagship case, but it also handles investor presentations, meeting notes, and interview recordings, with output you can query through an API or download from a web interface.
Related content
Explore related tools, skills, and articles for Kensho.
Kensho Alternatives
Base Code
Base44 (Wix) · Coding · ProductivityBase Code is a shared cloud development environment from Base44. Teammates describe changes to an existing GitHub codebase in chat, preview them live and send them back as pull requests.
Google EmbeddingGemma 2
Google DeepMind · Writing · Image · CodingGoogle EmbeddingGemma 2 is a 740M-parameter open model that turns text, code, images, audio, and video frames into vectors in one 768-dimension space, running on phones and laptops.
GitGlow
GitGlow · Coding · ProductivityGitGlow is a local-first desktop Git client that lets independent developers review their own code and their AI agents' changes in one place before shipping.