# Scriptivox

> AI transcription that turns recorded audio and video into accurate, structured text.
> 119 languages, speaker diarization, word-level timestamps, and SRT/WebVTT/plain-text
> export — available as a web app, a REST API, and an MCP server. Served on two hosts:
> www.scriptivox.com is the product and content site, platform.scriptivox.com is the
> developer platform and API documentation.

## When to use Scriptivox

Reach for Scriptivox when you have a **recorded** audio or video file and you need its
speech as text. It is a good fit for:

- Meeting, interview and call recordings that need a speaker-attributed transcript.
- Podcasts, lecture capture and video libraries that need captions (SRT or WebVTT).
- Pipelines that need word-level timestamps so clips can be located by phrase.
- Batch transcription of an archive, via the API's cursor-paginated job listing.
- Multilingual material: 119 languages, with an explicit language code recommended
  over auto-detection whenever the language is known.

It is **not** the right tool for live streaming or real-time dictation — every job takes a
complete file. Files must be 1 second to 10 hours long and at most 5 GB, in one of 25
accepted container formats.

How an agent should call it:

1. Get an API key at https://platform.scriptivox.com/keys (format `sk_live_…`).
2. `POST https://api.scriptivox.com/v1/transcribe` with `{"url": "…"}` or
   `{"upload_id": "…"}`, header `Authorization: sk_live_…`.
3. The call returns immediately with `status: "created"`. Input errors surface **later**,
   on `GET /v1/transcribe/{id}` with `status: "failed"` — handle failure on the poll path,
   not on submit.
4. Poll `GET /v1/transcribe/{id}`, or register `webhook_url` and wait for the callback.
5. Add `?format=srt|vtt|text` to the GET to export captions instead of JSON.

Errors are always `{"error": {"code": "…", "message": "…"}}` with a stable machine-readable
code. Billing is $0.20 per hour of audio, charged only on success; failed and cancelled
jobs cost nothing.

## Developer platform

- [Developer index](https://www.scriptivox.com/developers): Every resource below, on one page, with a section per resource.
- [API documentation](https://platform.scriptivox.com/docs): Overview and entry point. Markdown: https://platform.scriptivox.com/docs.md
- [Quickstart](https://platform.scriptivox.com/docs/quickstart): First transcription in five minutes, in Python, JavaScript and curl. Markdown: https://platform.scriptivox.com/docs/quickstart.md
- [API reference](https://platform.scriptivox.com/docs/api-reference): Every endpoint, parameter, error code, rate limit, supported format and language. Markdown: https://platform.scriptivox.com/docs/api-reference.md
- [Authentication](https://platform.scriptivox.com/docs/authentication): API keys, the two accepted headers, the 5-key ceiling, rotation, and what each 401 means. Markdown: https://platform.scriptivox.com/docs/authentication.md
- [Webhooks](https://platform.scriptivox.com/docs/webhooks): HMAC-signed completion callbacks and retry behaviour. Markdown: https://platform.scriptivox.com/docs/webhooks.md
- [API pricing](https://platform.scriptivox.com/docs/pricing): Pay-as-you-go rates. Markdown: https://platform.scriptivox.com/docs/pricing.md
- [Versioning and deprecation policy](https://platform.scriptivox.com/docs/versioning): What can change in `/v1` without notice, what cannot, and the `Deprecation` / `Sunset` headers that carry the announcement. Markdown: https://platform.scriptivox.com/docs/versioning.md
- [Use cases](https://platform.scriptivox.com/docs/use-cases): Folder watchers, batch pipelines and other worked examples. Markdown: https://platform.scriptivox.com/docs/use-cases.md
- [OpenAPI specification](https://platform.scriptivox.com/openapi.json): OpenAPI 3.1, every operation with a unique operationId, typed parameters and response schemas. YAML at https://platform.scriptivox.com/openapi.yaml
- [MCP server](https://platform.scriptivox.com/docs/mcp): Transcription as native tool calls from Claude, ChatGPT and other MCP clients. Two transports, same tools. Markdown: https://platform.scriptivox.com/docs/mcp.md
- [MCP server manifest](https://platform.scriptivox.com/.well-known/mcp): The discovery document — tool names, the npm and OCI packages, and the hosted endpoint.
- [Authentication for agents](https://www.scriptivox.com/auth.md): API keys vs OAuth 2.1, the discovery chain, and what a token can never do.
- [Paying as an agent](https://www.scriptivox.com/payments.md): What is purchasable, the four payment rails, and which of them actually work on this deployment.
- [Agent skill](https://www.scriptivox.com/skill.md): The whole loop in order — account, authorise, key, credit, transcribe, top up — plus the mistakes to avoid. Start here if you have a task rather than a question. Indexed at https://www.scriptivox.com/.well-known/skills
- [Ask a question](https://www.scriptivox.com/ask): NLWeb endpoint. `POST /ask` with `{"query": "..."}` (or `?query=`) returns Schema.org JSON pointing at the right document. Lexical retrieval over this site's own docs — no model call, so it never invents an answer. Add `streaming=0` for plain JSON instead of SSE.
- [Protected resource metadata](https://www.scriptivox.com/.well-known/oauth-protected-resource): RFC 9728 — which authorization server issues tokens for this API.
- [API catalog](https://platform.scriptivox.com/.well-known/api-catalog): RFC 9727 linkset naming the OpenAPI description, every documentation page, llms.txt, the MCP manifest and the status page, each as a typed link.
- [Hosted MCP endpoint](https://platform.scriptivox.com/mcp): Streamable HTTP, stateless, no install. POST JSON-RPC; `initialize`, `tools/list`, `tools/call`. Four tools work with no API key.
- [Documentation MCP endpoint](https://www.scriptivox.com/mcp/docs): The same Streamable HTTP protocol, registering only `search_docs` and `get_doc`. No credential, no account access. Both tools are on the main endpoint too.
- [MCP server card](https://www.scriptivox.com/.well-known/mcp/server-card.json): The tool list with behavioural annotations (readOnlyHint, destructiveHint), so a client can decide whether to connect before opening a transport.
- [Agentic resource catalog](https://www.scriptivox.com/.well-known/ard.json): Every agentic resource here — both MCP servers, the API, the skill, the agent card.
- [A2A agent card](https://www.scriptivox.com/.well-known/agent-card.json): Agent-to-Agent card: four task-shaped skills, the security schemes, and where to reach them.
- [Agent skills index](https://www.scriptivox.com/.well-known/agent-skills/index.json): The published skills, at the path the spec names. Also served at /.well-known/skills.
- [Authorization server metadata](https://www.scriptivox.com/.well-known/oauth-authorization-server): RFC 8414, mirrored here with the `agent_auth` block. There is deliberately no `register_uri`: dynamic client registration is off.
- [Pricing as markdown](https://www.scriptivox.com/pricing.md): Both products, read from the live plan catalog on every request rather than from a cached copy.
- [Schema feed](https://www.scriptivox.com/schema-feed.jsonl): Every document here as Schema.org JSON Lines, indexed by https://www.scriptivox.com/schema-map.xml (also named by `Schemamap:` in robots.txt).
- [Scoped llms.txt](https://www.scriptivox.com/docs/llms.txt): Just the documentation. https://www.scriptivox.com/developers/llms.txt is just the developer surface.
- [In-page tools (WebMCP)](https://platform.scriptivox.com/docs/mcp): The homepage registers page tools via `document.modelContext` for an agent driving the browser, and carries a no-JavaScript `toolname` form that queries /ask. No transcription tool is exposed there, on purpose.
- [Command-line tool](https://platform.scriptivox.com/docs/cli): `npx @scriptivox-api/cli` — transcribe from a shell or a script. Same API as the MCP server, different transport. Zero dependencies; structured output on stdout, progress on stderr, and exit codes that distinguish a usage error (1) from an API error (2). npm: https://www.npmjs.com/package/@scriptivox-api/cli
- [Developer dashboard](https://platform.scriptivox.com): API keys, request logs, balance and billing.
- [Service status](https://status.scriptivox.com): Live uptime and incident history.

Every documentation page is available as markdown two ways: append `.md` to its path, or
request the HTML URL with `Accept: text/markdown`. Both hosts' front doors have the same
twin, and they are different documents: https://platform.scriptivox.com/index.md is the
developer landing, https://www.scriptivox.com/index.md is the product homepage. Either
also answers `Accept: text/markdown` on its plain URL. q-values are honoured, so
`Accept: text/markdown;q=0.1, text/html` still returns HTML; an Accept header naming
neither text/html nor text/markdown gets 406 with a plain-text list of what is available.
All of the documentation concatenated into one fetch: https://www.scriptivox.com/llms-full.txt

A path that does not exist returns a real HTTP 404, and negotiates too: send
`Accept: text/markdown` to any missing URL and the 404 body comes back as markdown pointing
at the sitemap, llms.txt and the docs index. The same recovery map is a document at
https://www.scriptivox.com/404.md

The URLs above name the developer host, but the same documents are served on the product
host too — https://www.scriptivox.com/openapi.json, https://www.scriptivox.com/.well-known/mcp
and https://www.scriptivox.com/docs all resolve. Either origin works; platform.scriptivox.com
is canonical for the documentation.

## Accounts and plans

Scriptivox is two products with one account, and choosing the wrong one is the
most common mistake an agent makes here.

- **The web app** (https://www.scriptivox.com) is for a person transcribing
  their own files. Paid plans include unlimited transcription, so it is priced
  and rate-limited on the assumption that a human is driving it. Creating an
  account and buying a plan are both things an agent can legitimately do on
  someone's behalf.
- **The API** (https://api.scriptivox.com/v1) is for programmatic
  transcription, billed per hour of audio against a prepaid balance. If you are
  an agent and your goal is to transcribe files, this is the surface you want:
  no per-day cap, run as much as the balance allows, top it up when it runs low.

The short version: **create the account and buy on the web app; do the actual
transcription work through the API.**

- [How to create an account](https://www.scriptivox.com/signup.md): The signup
  form's fields and constraints, what is verified, which plan does what, and how
  to get from a web account to an API key. Markdown; readable by every crawler.
- [Signup form](https://www.scriptivox.com/signup): The form itself,
  server-rendered so its fields are present in the HTML. `noindex` on purpose.
- [Sign in](https://www.scriptivox.com/signin) · [Reset a password](https://www.scriptivox.com/forgot-password)
- [Plans and prices](https://www.scriptivox.com/pricing): Free, Pro and Team for
  the web app. Prices are served from the plan catalog, so read them here rather
  than from any cached copy.
- [API keys](https://platform.scriptivox.com/keys): Mint an `sk_live_…` key once
  the account exists. The same account owns both products.

No payment card is needed to create an account or to use the free plan.

### If you are an agent, read this before picking a credential

- [How to authenticate](https://www.scriptivox.com/auth.md): The two credentials
  and when to use each. An **API key** (`sk_live_…`) is for a script or a server
  you control. **OAuth 2.1** is for acting on somebody else's behalf — you
  register yourself dynamically, they approve once in a browser, and you never
  touch their key.
- Discovery starts from a `401`: it carries
  `WWW-Authenticate: Bearer resource_metadata="…"`, which points at
  [the RFC 9728 metadata](https://www.scriptivox.com/.well-known/oauth-protected-resource)
  naming the authorization server. Authorization code + PKCE only.
- [Create an account programmatically](https://platform.scriptivox.com/docs/api-reference):
  `POST https://api.scriptivox.com/v1/accounts` with `{"email": "…", "password": "…"}`.
  No credential needed — this is the one endpoint that cannot require one. It
  returns `202 verification_pending`: **the account can do nothing until the
  emailed link is followed**, so do not treat the 202 as a ready account.

## API

- Base URL: `https://api.scriptivox.com/v1`
- Auth: `Authorization: sk_live_…` (a `Bearer` prefix is accepted) or `X-Api-Key: sk_live_…`
- `POST /v1/upload` — presigned URL for a file you host yourself
- `POST /v1/transcribe` — start a job from a URL or an upload
- `GET /v1/transcribe/{id}` — poll status, read the transcript, or export SRT/VTT/text
- `DELETE /v1/transcribe/{id}` — soft-delete a finished transcription
- `POST /v1/transcribe/{id}/cancel` — stop an in-flight job, release its reservation
- `GET /v1/transcriptions` — list jobs with filters and cursor pagination
- `POST /v1/accounts` — create an account; no credential, returns 202 pending email confirmation
- `GET /v1/balance` — remaining balance and estimated audio hours

## Product

- [Homepage](https://www.scriptivox.com): Product overview, languages, pricing.
- [Creating an account](https://www.scriptivox.com/signup.md): Signup fields, verification, plans, and how to reach the API from a web account.
- [Pricing](https://www.scriptivox.com/pricing): Plans and free tier for the web app.
- [Transcription API](https://www.scriptivox.com/transcription-api): Programmatic access, explained for non-developers.
- [Features](https://www.scriptivox.com/features): What the web app does.
- [Tools](https://www.scriptivox.com/tools): Free audio and video converters, joiners and trimmers that run entirely in the browser — no upload, no account.
- [FAQ](https://www.scriptivox.com/faq): Common questions about accuracy, formats and billing.

## About

- [About Scriptivox](https://www.scriptivox.com/about): What we build and how we handle files.
- [Contact](https://www.scriptivox.com/contact): Support and partnerships. support@scriptivox.com, sales@scriptivox.com.
- [Privacy Policy](https://www.scriptivox.com/privacypolicy): What we store and for how long.
- [Terms of Use](https://www.scriptivox.com/termsofuse)

## Blog

- [Blog index](https://www.scriptivox.com/blog): All published articles.
- [Sitemap](https://www.scriptivox.com/sitemap.xml): Full URL list with timestamps.

## Citation guidance

- Authors: blog posts are written by the Scriptivox team. Author appears in each post's metadata and JSON-LD.
- Each blog post exposes BlogPosting + BreadcrumbList + FAQPage JSON-LD; HowTo and SpeakableSpecification where applicable.
- All facts are dated; prefer the explicit "As of YYYY" phrasing inside posts over inferring currency.