Captions.ai alternative for developers: an API that captions videos in your pipeline

Captions.ai is one of the best-known names in AI captioning, and it earns it: a creator opens the app, records or uploads a talking-head video, and gets captions, cuts, and polish with almost no effort. If you make one video at a time and want a GUI, it is a good tool. This post is for everyone else: teams whose videos are produced by other software, and who keep searching for a Captions.ai alternative because an editor is the wrong shape for a pipeline.

The difference is the interface, not the quality

An editor assumes a human in the loop. Every video passes through a person who imports it, checks the captions, and exports. That is a feature when the video is your product. It is a bottleneck when the videos come out of a clipping tool, a UGC ad generator, a podcast automation, or an agent that posts on a schedule. Software that makes videos needs captioning to be one more function call, not a tab someone has to open.

caption.sh is built for that case. There is no editor to learn because there is nothing to edit by hand: you send a video, the API returns a captioned MP4. Speech-to-text with word-level timing, caption grouping, styling, and rendering all happen server-side.

What the API looks like

One POST creates a job, you poll or take a webhook, and you download the result:

curl -X POST https://api.caption.sh/v1/videos \
  -H "Authorization: Bearer $CAPTION_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"source_url": "https://example.com/raw.mp4", "options": {"max_words": 2}}'

# -> 202 { "job_id": "..." }
# poll GET /v1/jobs/{id} until status is "done"
# then GET /v1/jobs/{id}/result for the captioned MP4

The options object carries the style: max_words for the karaoke look, highlight_color, decoration, position, font. Font size is a fraction of video height, so the same JSON produces proportionally identical captions on a vertical Short and a landscape clip. 18 fonts are bundled, so there is nothing to install or license.

If your pipeline is an agent

caption.sh also runs a hosted MCP server at https://mcp.caption.sh/mcp. Claude Code, Cursor, and other MCP clients get caption_video and the job tools as native calls, plus nine style recipes, so an agent can caption a folder of clips from one sentence. An editor cannot be driven that way, and that is the whole point: when the caller is software, the interface is the product.

Pricing, honestly

Editors sell subscriptions: a flat monthly fee whether you caption two videos or two hundred. caption.sh is prepaid usage at $0.10 per minute of video, charged per second, only for successful renders. A hundred one-minute clips cost $10 with no plan and no seat. If you caption three videos a month and like a GUI, a subscription editor is fine. If you caption thousands, or the caller is software, usage pricing wins by construction.

When Captions.ai is still the right call

If you are a creator recording yourself, want AI editing beyond captions (cuts, zooms, B-roll), and enjoy working in an app, use the app. We are not trying to replace an editor. We are the thing you pick when the editor is the bottleneck. The two products answer different questions that happen to share the word captions.

Try it

The playground on caption.sh shows the output in seconds: drop a clip, tune the style, watch the preview. API keys are in the dashboard, the reference is at caption.sh/docs, and the MCP server takes the same key as a bearer token.