Free Index-Translate API for client subtitles: what to check

Index-Translate's free endpoint is labelled an online demo with no stated rate limit. A checklist before client transcripts go in, and the self-host option.

5 min readSume
All posts

Use the free Index-Translate endpoint for public or low-stakes subtitles, and self-host the open weights before sending anything a client considers confidential. The page that advertises the endpoint calls it free and OpenAI-compatible, labels it an online demo, and publishes no rate limits, retention policy or service commitment that we could find on the pages read on 2026-10-04.

That does not make it unsafe, only unspecified. For a client transcript, unspecified is the thing to resolve first.

What does the vendor say about the free endpoint?

Bilibili's Index team released the Index-Translate family on September 30, 2026. The weights are Apache-2.0, which allows you to run them yourself. The hosted endpoint is a separate convenience on top of that.

Hosted endpoint versus open weights, read 2026-10-04
QuestionFree hosted endpointSelf-hosted weights
Cost statedFreeYour own compute
Rate limitsNot documented on the pageWhatever your hardware serves
Data stays with youNot statedYes, if you run them in your environment
LicenceNot stated on the pageApache-2.0
Uptime commitmentNone stated; labelled an online demoYours to provide

What should you check before sending a transcript?

If any answer is no, run the 2B or 9B weights yourself. Both sizes are in the repository, and the burn step does not care where the text came from.

  • Does your client contract allow third-party processing of the spoken content, even for a few lines?
  • Is the transcript already public, such as a published video? Then the privacy question mostly disappears.
  • Can you strip names, numbers and addresses from the cues first and restore them after translation?
  • Do you have a fallback if the demo endpoint is slow or gone mid-project?

What does the Sume side see?

Sume's video captions endpoint receives two things: a public HTTPS video_url and your cues. Media inputs says localhost, private-network, non-HTTPS and signed or private URLs are rejected, so the clip itself has to be fetchable. That is a different exposure from the transcript sent to a translation endpoint, and worth deciding separately.

A standalone caption job is priced at $0.20 for videos up to 60 seconds under the current estimate. Cues are capped at 200 per request, 400 characters each, within 60 seconds.

A workable policy

Use the free endpoint for your own channel's content and for pilots. For client work, either self-host or get written approval for the hosted endpoint, and keep a record of which one translated which cues. The cue-translation recipe in this walkthrough works the same with either source of translated text.

Sources

Related posts

More in Developers

All Developers posts

Written by Sume