A video API built for text, not for prompts

The short answer

If your input is written content, an article, a report, a newsletter issue, a PDF or a slide deck, and your output needs to be a narrated video that matches your brand, the Blog2Video API is built for exactly that job. You send a link or the text. It reads the structure, writes a scene by scene script, records the voiceover, and renders an MP4 in a template.

If you want cinematic footage from a prompt, or a talking human presenter, it is not the right tool, and this post says so plainly below. The rest of this page is about where it is the best choice, and why.

Four calls from article to MP4

The whole happy path is four requests, all authenticated with a bearer API key sent from your server.

• POST /api/v1/videos with a url or content starts generation and returns a video id.
• GET /api/v1/videos/{id}/status tells you how far generation has got. Poll it until the script and scenes are ready.
• POST /api/v1/videos/{id}/render renders the video.
• GET /api/v1/videos/{id}/render returns the render status; when it reports done, video_url is the link to your finished MP4.

Options that matter when you create a video

Only url or content is required. Everything else has a sensible default, so a first request can be one line. When you want control, you can pick a template from the catalog or point at one of your own custom templates, choose landscape or portrait, and set the voice gender or turn narration off.

GET /api/v1/catalog lists the built-in templates, so a picker in your own product can be built from the same source the Blog2Video app uses.

• template: an id from the catalog, or custom_<id> for a template built from your own brand
• aspect_ratio: landscape or portrait
• voice_gender: female, male or none
• external_user_id and idempotency_key: tag a video with your own user id, and make retries safe

Beyond a URL: documents, editing and translation

The project API behind the simple endpoints is much deeper than four calls. You can create a project from uploaded PDF, DOCX, PPTX, TXT, Markdown or VTT files, up to five files of 5 MB each. You can regenerate a single scene, add a scene from a prompt, reorder scenes, swap images, set a stock clip, change the voice, upload your own voiceover for a scene, and translate a whole project, which rewrites the text and re-records the narration.

Custom voices are supported too, including voice design from a prompt. Cloning a voice from a recording needs a paid plan.

• Script review: create a project with script_review_enabled and generation pauses so a person can approve the script before anything is rendered.
• Scene level edits: regenerate, add, reorder and restyle without rebuilding the video.
• Brand: upload a logo, use a custom template, keep every video on the same look.

Built to sit inside someone else's product

Newsletter tools, Ghost and WordPress hosts, finance media companies and agent builders would all rather embed video than build a render pipeline. The API is designed to be called from a backend: keep the key on your server, create videos on behalf of your users, and tag each one with an external_user_id so you can list a single user's videos later with the same id as a filter.

One honest note for platform builders. A key belongs to one Blog2Video account, so your own backend is responsible for deciding which of your users may see or edit which video. Notestack, which is built this way, checks ownership on every route before it forwards a request. The case study walks through how.

Read how Notestack uses the Blog2Video API

Agents can call it too

Blog2Video also ships an MCP server, so an AI assistant or workflow tool can create and monitor videos without you writing any glue code. There are setup guides for Claude, ChatGPT, Gemini, DeepSeek and n8n. Every agent that connects is a user you did not have to onboard.

Set up the MCP server with Claude

Why not a generative video API?

Generative video models are impressive and they are the wrong tool for a report with numbers in it. They draw pixels, so text can come out garbled and a figure can quietly change between frames. A template based renderer draws real text and real charts from the content you give it, in the same layouts every time.

That difference is why finance and data heavy publishers are the natural users of this API. The companion post covers it in detail.

Why data heavy publishers choose template based video

Where it is not the best choice

No tool is best at everything, and you should know the limits before you build on it.

• No AI avatars or talking human presenters. If a face is the product, use an avatar platform.
• No photoreal footage generated from a prompt. Scenes are built from your text, images and optional stock clips.
• No webhooks today. You poll the status endpoints until generation and rendering finish.
• The API needs a paid plan, and each video counts against your plan's monthly allowance.
• The script is written by an AI model, so wording can differ between runs. For anything high stakes, use script review and read it before you render.

Pricing and getting started

API access is on paid plans, which start at 9.99 dollars a month for 10 videos, with higher tiers for 30 and 100 videos a month. Create a key on the API Keys page, export it as an environment variable, and run the first request from the docs. You can have a finished MP4 in minutes.

Open the API documentation

Frequently Asked Questions