Sonix SRT and VTT export vs Sume burned-in captions

If you need a subtitle file you can upload to a platform, Sonix is built for it and Sume video captions are not. Sonix's features page (read 2026-10-10) lists SRT and VTT export plus burned-in subtitles, while Sume's caption job burns styled words onto your video and returns a captioned video_url.

The Sume docs say the resource returns the status, the style and the captioned video, and that raw transcripts are not part of the public contract. That is a clear boundary, so the choice is about which output your platform needs.

Output and pricing at a glance

Sonix sells transcription by the hour with monthly plans; Sume sells a fixed caption job. Sonix's page calls out 54+ languages and "99% accuracy"; Sume's caption docs describe a language hint for speech-to-text and do not state a language count.

What a Sume caption job does well

The strength is the look. Standalone captions take a style such as slam (the Latin default), punch or tiktok-green, plus Hangul styles for Korean speech. A design object changes colors, typography, placement, phrasing and motion for one request, and out-of-range values fail with a 400 at request time rather than producing a bad render you pay for.

If you hold the exact wording, send script_text to lock the burned-in text to your script while speech-to-text timings stay the timing source. If alignment cannot map the script onto the speech, the job fails with script_alignment_mismatch or script_alignment_failed. For a clip with no speech, pass cues instead; otherwise the job fails with caption_no_speech.

The gap: no sidecar file

Platforms that accept a separate subtitle track, so viewers can toggle it, want SRT or VTT. Sume's caption docs do not describe producing one, and say the API does not support SRT uploads either. A burned-in caption is permanent and cannot be turned off by the viewer or translated by the platform.

One path exists on the Sume side: speech-to-text always returns word timings, with optional sentence segments, so you could build a subtitle file from that output in your own code. I am not claiming Sume ships that as a feature, only that the timing data is in the STT result.

Choosing

Use a short list.

• Viewers should toggle or translate captions: you need SRT or VTT, so Sonix.
• Short vertical video where captions are part of the design: Sume captions at $0.20 per clip.
• Both: transcribe and export a file elsewhere, and burn the styled version with Sume.
• Clips longer than 60 seconds: the $0.20 figure covers up to 60 seconds under the current estimate, so check the live rate before batching long videos.