How do punctuation and line breaks differ between subtitles and transcripts?
Subanana formats text differently depending on the processing mode you choose. This article explains how subtitles (subtitle generation) and transcripts (transcript / AI meeting summary) handle punctuation and line breaks, and how you can adjust the results yourself.
1. Different modes format text differently
Transcript (transcript mode): the system outputs clear, readable text and supports speaker identification (diarization). Best for interviews, conversations, research recordings, and other content you need to read in full.
AI meeting summary: on top of the formatted transcript, an AI summary is added that pulls out key points, decisions, and action items.
Subtitle generation: the system automatically converts your video/audio into multilingual subtitles, breaking lines so each subtitle stays at an easy-to-read length rather than presenting full written paragraphs.
Transcript and AI meeting summary modes automatically run a formatting step that groups continuous speech into paragraphs and restores punctuation, making the text easier to read. This step is always on and cannot be toggled manually (nor does it need to be). Subtitle generation does not run this paragraph-grouping and punctuation-restoration step; it focuses on subtitle line breaks instead.
As a result, the same audio will look different in subtitle mode and transcript mode (including line breaks, punctuation, and text formatting). This is a design difference between the modes, not an error.
2. Why do subtitles sometimes show two sentences merged into one long line?
After transcription, subtitle generation runs an AI sentence-splitting and line-breaking process that cuts overly long sentences into subtitle lines of a comfortable reading length.
In most cases, this process automatically splits long sentences into multiple lines.
However, this process is probabilistic, and line breaking can occasionally fail for individual sentences. When that happens, the system falls back to showing the original, unsplit result. That is why you sometimes see what should be two sentences displayed as one.
How to fix subtitles that are too long or not split
Re-transcribe: in the editor for that video, use the rating prompt and choose "Not ideal" → "Subtitles missed a lot of content". The system will transcribe the file again with more suitable settings and run the line-breaking process again. This retry does not use any extra minutes, and in most cases it improves the result.
Edit manually in the editor: if individual sentences are still not right after a retry, you can split that line yourself in the Subanana editor (or merge two adjacent lines).
Note that line breaking is probabilistic: a retry has a real chance of improving the result, but it is not guaranteed to fix a specific sentence every time.
3. Is there a setting to limit the number of characters per subtitle line?
There is currently no option to set a maximum character count per subtitle line or a duration per line. Subtitle line breaks are handled by the AI process described above. To control the length of a specific subtitle line, split or merge that line yourself in the editor.
4. Will punctuation be added automatically to an uploaded SRT file?
No. Punctuation restoration is a formatting step that transcript and AI meeting summary modes run when transcribing from audio; it is not applied to subtitle files you upload. The system does not re-punctuate user-uploaded SRT files. To adjust punctuation in an existing subtitle file, edit it yourself in the editor.
5. Paid plans and transcription engines
On the Free plan, only the first 15 minutes of each file are generated (for file modes such as subtitle generation, transcript, and AI meeting summary); the rest is not processed. After upgrading, you can process the remainder of the same file in the editor via "Continue generating subtitles / Continue generating transcript". Credits are deducted only for the remaining duration, and there is no need to re-upload. (The free trial of real-time subtitle translation is one-time, with a separate 5-minute limit — it does not renew.)
If transcription quality for the same content is repeatedly unsatisfactory (including missed content or sentence formatting), you can choose the "High accuracy" transcription engine when creating the project. Its accuracy may be higher than the standard engine's; this option is billed at 1.5x credits (video duration × 1.5).
If the above does not solve your problem, feel free to contact us in the app or email hello@subanana.com. We reply within one business day (Monday to Friday, Hong Kong time).