Transcription API plus the AI platform built on top of it.

Transcription API with AI platform: why they belong together

Most transcription APIs are point solutions. You get an endpoint, a response, and a transcript. What you do with that transcript is entirely your problem. You need a separate LLM call for summarization, a separate database to store and search transcripts, a separate UI for your non-technical teammates, and a separate billing relationship for every layer you add. Speak is built on the premise that those layers belong together in one account, under one pay-as-you-go credit balance, with no subscription required to get started.

Що означає оплата за використання для розробників API

Pay-as-you-go on Speak means a prepaid credit balance that you top up in any amount. There is no monthly subscription charge, no minimum monthly spend, and no expiration date on credits you add. Transcription is metered per second so a 4-minute call costs roughly the same fraction of an hour as it takes. The AI Meeting Assistant and the transcription API share the same credit rate. AI Chat over your transcripts is billed in credits, based on usage. All of this pulls from one credit balance regardless of which surface you use: the REST API, the CLI, the MCP server, or the dashboard.

For developers this means you can onboard without a procurement cycle, test a full pipeline without committing to a subscription tier, and scale usage up or down without renegotiating. If you stop using Speak for two months, your credit balance is waiting for you when you come back.

How the API fits into the broader platform

The Speak REST API is the foundation. Every other surface in the product is built on the same endpoints. The CLI wraps the upload, search, export, and analyze endpoints in 26 terminal commands. The MCP server exposes 83 tools to Claude, Cursor, Windsurf, and any MCP-aware AI client. The web and mobile UI lets non-technical teammates use the same transcription and AI analysis that your code triggers via API. One Speak account covers all four surfaces. One pay-as-you-go credit balance covers all usage.

Transcription API plus AI Chat: the compound value

Transcription gets you a text representation of audio. AI Chat over that transcript gets you answers. Ask a question across one file or your entire library. Identify speakers, extract decisions, find every call where a competitor was mentioned, or summarize the 50 longest interviews in your corpus. AI Chat is billed in credits, based on usage, model routing is handled for you, and the same pay-as-you-go credit balance covers it. No second account, no second API key, no second billing relationship.

MCP server for Claude and other AI clients

Speak MCP сервер дає AI помічникам прямий доступ до вашого робочого простору Speak. Claude Desktop, Claude Code, Cursor та Windsurf можуть транскрибувати нові аудіо, шукати існуючі транскрипції, запускати Magic Prompts, витягувати сегменти спікерів, експортувати вміст та багато іншого, все через природну мову. Встановіть MCP сервер з npm, вставте свій Speak API ключ, і будь-який MCP-сумісний клієнт матиме 83 доступні інструменти. Той же залишок кредитів з оплатою за використання охоплює все використання, викликане MCP. Додаткова конфігурація виставлення рахунків не потрібна.

Getting started with the Speak transcription API

Sign up for a free account at app.speakai.co. Open the Developers section of your dashboard to generate an API key. No credit card required. The 7-day trial includes full API access. When you are ready to scale, top up your pay-as-you-go credit balance in any amount. Credits never expire. Full API reference, SDK guides, and webhook documentation are at docs.speakai.co.

Webhooks and async analysis

Speak fires a signed webhook when transcription and AI analysis are complete. Register your endpoint via POST /v1/webhook and specify which event types you want. The media.analyzed подія спрацьовує, коли транскрипція та всі AI інсайти готові. Використовуйте її для активації автоматизації, що вниз по потоку: помістіть транскрипції в ваш CRM, запустіть Slack сповіщення, оновіть рядок бази даних або запустіть користувацький Magic Prompt для кожного нового запису. Webhooks є частиною того ж облікового запису з оплатою за використання, того ж API ключа, того ж залишку кредитів.

Speaker diarization and multi-language support

The Speak transcription API supports 100+ languages and automatic language detection. You can specify sourceLanguage on upload or let Speak identify the language for you. Speaker diarization labels each turn in a conversation by speaker, making transcripts of interviews, calls, and meetings immediately parseable without manual cleanup. Word-level timestamps are available on every transcript so you can build video chaptering, subtitle files, or searchable moment-in-time interfaces on top of the API response.

Порівняння постачальників API з оплатою за використання у 2026 році

The pay-as-you-go transcription API market has three meaningful options in 2026: AssemblyAI, Deepgram, and Speak. AssemblyAI is the strongest choice if you need an API-only workflow with built-in LLM tooling (LeMUR) but no end-user workspace. Deepgram has the lowest per-hour rate for raw transcription and strong real-time streaming support, but there is no AI analysis layer or workspace UI. Speak is the right choice when you need transcription plus AI Chat plus a workspace UI that non-technical teammates can use, all under one pay-as-you-go credit balance with no subscription required. The empty quadrant in 2026 belongs to Speak: the only vendor that ships all four layers in a single account.

Using the Speak CLI and MCP server from your API key

Your Speak API key works across every surface. The CLI lets you transcribe a local file with one command: speak transcribe interview.mp3. Той же API ключ та той же залишок кредитів з оплатою за використання. MCP сервер використовує той же ключ для доступу Claude та Cursor. Вам ніколи не потрібно керувати кількома ключами, кількома обліковими записами виставлення рахунків або кількома наборами облікових даних. Один обліковий запис Speak, один ключ, один залишок, кожна поверхня.