Instagram voice translation covers nine languages; add a tenth by cues

Instagram's voice translation covers nine languages, so a Reel in any tenth language needs another route. The Meta newsroom post (read 2026-10-05) lists English, Hindi, Spanish and Portuguese as available earlier, and adds Bengali, Tamil, Telugu, Kannada and Marathi. For a market outside those nine, the practical option is to write the translated lines yourself and burn them as caption cues.

This post is about that fallback. It does not claim that Sume can dub a voice. It covers the part Sume does: placing your own translated text on a finished video.

What the Meta post says

According to the post, the tool keeps the sound and tone of the creator's voice, and creators can switch on lip sync so the translated audio follows the mouth. The post says voice translations were rolling out to all users as of January 16, 2026, and adds two font options, Devanagari for Hindi and Marathi, and Bengali-Assamese for Bengali and Assamese, with the fonts reaching Android in the coming days. The announcement was made at an event in Mumbai.

Everything above is the vendor's description, and availability can vary by account and country, so check the app before you build a plan around it.

The caption fallback on Sume

Suppose your Reel is in one of those nine languages and you want a tenth market. First, get a clean video file with no text in it. Second, write the translated lines and their timings. Third, send them as cues to the Sume video captions job. With cues (or segments), Sume does not transcribe the audio, so the burned text is exactly what you wrote. That matters here, because the audio is in a different language from the text.

The caption job costs $0.20 for a clip of up to 60 seconds. The language field is only a hint for speech-to-text and never picks a style or font, so it does nothing for cues.

• Keep the clean master. Burned text cannot be removed from a file.
• Match phrase length to speech pace. The design.phrasing fields max_words, max_chars and pause_seconds give you control.
• Pick a face that covers your script. The Sume docs document Latin faces and Hangul faces, and say that Hangul text on a Latin style returns 400 caption_hangul_text_latin_style. They do not document faces for Devanagari or Bengali, so test a sample first.

Which route to choose

Use Instagram's tool when your language is in the nine and you want a translated voice. Use burned cues when you need text on screen, when your language is not in the nine, or when you need exact wording such as a legal line or a price. The two can be combined: translate the voice in the app, and keep a separate clean master for cue-based captions in the markets where muted viewing is common.

Whichever you pick, watch one finished Reel muted and once with sound before you publish. The wording on screen should match what is spoken, or the viewer will notice the gap.