# Subtitles that are truly readable vertically: characters per line, CPS, and minimum duration per cue

> Readability rules for subtitling vertical video: characters per line, 2 lines, reading speed in CPS, minimum duration per cue, and gaps.

- **Source (source):** https://aitoolsforapps.com/capcut-srt-subtitles-edit-plan-png-titles-245/blog/vertical-video-subtitles-characters-per-line-cps
- **Platform (platform):** AiToolsForApps
- **Language (language):** en
- **Updated (updated):** 2026-11-20

## In summary

- AI no longer fails in **what** is said, it fails in **how it is chunked**: the transcription gets the words right but leaves long lines, half-second cues, and three-line blocks over the speaker's face.
- The [BBC Subtitle Guidelines](https://www.bbc.co.uk/accessibility/forproducts/guides/subtitles/) (version 1.2.5, March 2026) set the vertical equivalent: what is 37 characters per line in a 16:9 is **about 25 characters** in a 9:16.
- Reading speed: the BBC recommends **160-180 words per minute** and a minimum of **0.3 seconds per word**; [Netflix](https://partnerhelp.netflixstudios.com/hc/en-us/articles/217350977-English-Timed-Text-Style-Guide) allows up to **20 CPS** for adults and **17 CPS** for children's content. For vertical, 12-17 CPS is the comfortable range.
- Duration per cue: Netflix requires a **minimum of 20 frames (4/5 of a second)** and a **minimum gap of 2 frames**; the BBC asks that, if there is a pause, the gap be **at least one second**.
- Since **June 28, 2025**, the [European Accessibility Act](https://eur-lex.europa.eu/eli/dir/2019/882/oj) is applicable, and in Spain, it is transposed by the [Law 11/2023](https://www.boe.es/buscar/act.php?id=BOE-A-2023-11022): subtitles have ceased to be a marketing extra.

---

If you edit vertical video, 2026 has gifted you the easy part of subtitling and left the difficult part intact. Automatic transcription is already good; chunking remains a craft. This article gathers the readability rules used by broadcasters and platforms, translates them to the 9:16 format, and explains how to review an `.srt` before importing it into your editor. In the end, you will see where **aiCapCut** fits in, the tool from **AiToolsForApps** to [generate and re-chunk SRT subtitles for vertical video](/capcut-srt-subtitles-edit-plan-png-titles-245) without touching the timeline.

![Vertical mobile frame with a two-line short subtitle placed in the lower third, away from the face of the speaker](https://cdn.aistohire.com/i/3052-oyaf-bftv-atpf-gkai.webp)

## Automatic subtitles get the words right but fail at chunking

Voice recognition solves transcription; chunking —where each cue starts and ends, where the line breaks, how long it stays on screen— is an editorial decision that no engine makes for you. That’s why an automatic subtitle can have 100% of the words correct and still be unreadable.

The three mistakes that keep appearing in vertical:

1. **Lines too long.** The engine chunks based on audio pause, not box width. In 9:16 a phrase that fit comfortably in horizontal spills over or is reduced to an unreadable font size.
2. **Cues too short.** An isolated filler word ("yeah", "okay") generates a cue of 300 milliseconds that blinks and doesn't allow time to focus the view.
3. **Three or four stacked lines.** The block grows upward and ends up covering exactly what the viewer needs to see: the mouth of the speaker.

The [BBC guide](https://www.bbc.co.uk/accessibility/forproducts/guides/subtitles/) states it plainly in its positioning section: in vertical videos, it’s advisable to raise the subtitles slightly, but keep them in the lower third, "because faces are usually in the upper half of the screen, and placing the subtitles at the very bottom makes it difficult to read the text and see the person speaking".

![Comparison of two vertical frames: on the left, four long subtitle lines cover the face; on the right, two short lines in the lower third leave the face visible](https://cdn.aistohire.com/i/4340-mulo-kleo-tiri-eyxb.webp)

## In vertical, about 25 characters fit per line, not 37 or 42

The BBC provides the exact equivalence: "the equivalent of 37 characters in a region of 75% of the width of a 16:9 video is 25 characters in a region of 90% of the width of a 9:16 video". In other words, the vertical format takes away about a third of the text per line compared to the broadcast standard.

Where the circulating figures come from:

- **37 characters** is a teletext limit: of the 40 bytes available per row, at least 3 are spent on control codes, leaving 37 usable positions. It’s an infrastructure number, not a physiology one.
- **42 characters per line** is the cap of the [English Timed Text Style Guide from Netflix](https://partnerhelp.netflixstudios.com/hc/en-us/articles/217350977-English-Timed-Text-Style-Guide), designed for large screens and proportional fonts.
- **25 characters** is the honest translation of those caps to 9:16, according to the BBC itself.

The same guide warns of something that is worth remembering: controlling length only with a character counter "is a crude method that does not take into account the width of each letter or the font". A line of 30 characters filled with m's and w's takes up more space than one of 34 with i's and l's. The counter is a good guardrail, not a guarantee: check the result on screen.

In practice, for vertical social media video, it works well to set the maximum between **28 and 34 characters per line** and drop to 18-22 if you are going to use single-line subtitles like karaoke.

## Two lines is the practical ceiling in networks, although the BBC allows three vertically

Here is a real divergence worth knowing. Section 3.3 of the BBC guide recommends a maximum of **two lines** for landscape or square video (16:9, 4:3, 1:1) and **three lines** for vertical video (9:16), taking advantage of the higher framing. In social video, however, the third line almost always comes at a cost.

The reason is that in a 9:16 for social media, the usable height is not the height of the file: the application interface, the description text, the side buttons, and the bottom bar eat up part of the frame, and the subtitle block grows upwards, toward the face. The BBC itself nuances its permission: extra lines can be used "if you have the confidence that no important information from the image is hidden," and in the section on cumulative subtitles, it insists on not using them if they will cover mouths or other relevant visual elements.

The operational conclusion: **set the limit at 2 lines** and treat the third as a justified exception, not as a default resource. If it doesn't fit in two lines of 30 characters, the problem is not the number of lines: it's that that cue should be two cues.

For line breaks, the BBC asks to break at logical points —ideally at a punctuation mark— and, if there is no other option, to avoid separating article and noun, preposition and its phrase, conjunction and the phrase it introduces, pronoun and verb, or the parts of a compound verb. Splitting a word in half is the worst thing you can do to reading.

## 12-17 CPS is the comfortable speed range for reading vertically

Reading speed is measured in characters per second (CPS): you divide the characters of a cue by its duration. If a cue of 45 characters lasts 2 seconds, it is at 22.5 CPS and is too fast to read it all while also looking at the image.

The published references:

| Reference | Speed | Document and date |
|---|---|---|
| BBC | 160-180 words/minute; minimum ~0.3 s per word (1.2 s for a subtitle of 4 words) | [Subtitle Guidelines v1.2.5](https://www.bbc.co.uk/accessibility/forproducts/guides/subtitles/), March 2026 |
| Netflix (adults) | Up to 20 CPS | [English Timed Text Style Guide](https://partnerhelp.netflixstudios.com/hc/en-us/articles/217350977-English-Timed-Text-Style-Guide) |
| Netflix (children) | Up to 17 CPS | [English Timed Text Style Guide](https://partnerhelp.netflixstudios.com/hc/en-us/articles/217350977-English-Timed-Text-Style-Guide) |

These limits are designed for someone sitting in front of a screen, with their attention on the movie. Vertical video is consumed standing, in motion, with the screen half a meter away and competing with the rest of the feed. That’s why the sensible recommendation for 9:16 is to stay clearly below Netflix’s limit: **between 12 and 17 CPS**, with 15 as a starting point and 12 if your audience is broad or the vocabulary is technical.

Be careful with how the parameter behaves: CPS do not lengthen the audio, they determine **how much text can go inside a cue**. Dropping from 17 to 12 CPS does not slow down the video; it forces earlier cuts and generates more cues.

## Each cue needs a minimum, a maximum, and a clean gap

Three temporal parameters decide whether a subtitle is read or blinks: the minimum duration per cue, the maximum, and the gap between consecutive cues. The two most cited references do not match, and the difference is explainable.

![Timeline with subtitle blocks of different widths: the very narrow ones in red are cues that are too short, the green ones have comfortable duration, and the amber ones stay on screen too long](https://cdn.aistohire.com/i/5676-arfa-bzee-zbsl-zejr.webp)

| Parameter | BBC (Subtitle Guidelines v1.2.5, March 2026) | Netflix (Subtitle Timing Guidelines) | Starting value in vertical |
|---|---|---|---|
| Minimum duration per cue | ~0.3 s per word (1.2 s for 4 words) | 20 frames, that is 4/5 of a second | 0.8-1 s |
| Maximum duration per cue | Does not set a numerical limit; asks that the subtitle does not stay on screen beyond the end of speech, because it gets re-read | Does not set a limit; asks for "good judgment" so it does not drag on | 4-6 s |
| Gap between cues | If there is a pause, minimum 1 s, preferably 1.5 s | Minimum 2 frames; gaps of 3 to 11 frames are closed to 2; either 2 frames or half a second or more | 0.05-0.1 s between chained cues |
| Entry regarding audio | Must coincide with the start of speech | At the first audio frame or 1-2 frames from it | First audio frame |
| Exit regarding audio | Approximately coincides with the end of the speech segment | Ideally at least half a second after the end of the audio | End of speech + short margin |

The apparent contradiction of the gap is understood by reading the fine print. Netflix regulates the **technical gap** between chained cues: two frames, so the eye notices the change without the screen blinking; and expressly prohibits intermediate gaps (3-11 frames), which produce the nervous effect. The BBC regulates the **editorial gap**: if there is a real pause in the speech and you decide to leave the screen clean, that pause must last at least one second, because "anything shorter produces a very choppy effect."

Translated to your `.srt`: between cues that follow one another, leave a small and constant gap (on the order of 50-100 ms). If you are going to leave a visible silence, let it be truly long.

The BBC adds a rule that is often forgotten: **keep durations consistent**. If you give a certain length subtitle 3 seconds, do not give 4 to the next one of similar length. Irregularity tires more than speed.

## In 9:16 the subtitle goes in the lower third, not at the very bottom

Position is the other factor of readability, and it changes vertically compared to the television standard. The BBC guide recommends, for 9:16, to raise the block slightly above the lower edge but keep it in the lower third of the screen: at the very bottom it gets covered by the interface, and too high it covers the face.

![Vertical frame with marked areas: in red the lower strip and the right edge that covers the interface, in green the band of the lower third where it is advisable to place the subtitle](https://cdn.aistohire.com/i/6697-mqne-sqep-zsjj-zbvz.webp)

The size is also different. The BBC sets the line height at **7-8% of the video height** for 16:9, 4:3, and 1:1, and at **3.9-4.5%** for 9:16. This makes sense: vertically, the height of the frame is much greater in proportion, so a lower percentage gives a comparable body of text.

Two practical consequences:

- **Leave side and bottom margins.** The BBC online guide talks about occupying up to 90% of the width in 9:16; the remaining 10% is your buffer against the edges.
- **Check the area that each network covers.** The side buttons, the description, and the navigation bar of each application cover different parts of the frame. What looks perfect in your preview may be partially hidden in the feed.

## How to review and re-slice an .srt before importing it

An `.srt` is plain text and can be audited before touching the editor. The order of review that works is: first the validity of the file, then the readability of each cue, and only at the end the style.

A well-formed cue looks like this:

```
12
00:00:21,480 --> 00:00:24,120
What AI doesn't decide for you
is where to cut the phrase.
```

**Format review (what breaks the import):**

1. Correlative numbering from 1, with no jumps.
2. Milliseconds separated by **comma**, not by period: `00:00:21,480`. The period is for WebVTT, not for SRT.
3. Consistent line breaks (CRLF) and a blank line between cues.
4. Strictly increasing times and **no overlaps**: the end of one cue can never be later than the start of the next.

**Readability review (what breaks reading):**

1. No cue above your maximum characters per line.
2. No cue with more than 2 lines.
3. No cue above your CPS limit: `characters ÷ (end − start)`.
4. No cue below the minimum duration or above the maximum.
5. Small and consistent gaps between chained cues.
6. Line breaks at logical points, without separating articles and nouns or splitting words.

**Editorial review (what differentiates a good subtitle):**

1. Each cue, ideally, one sentence or a complete unit of meaning.
2. Edit instead of transcribing literally when it doesn't fit: remove filler words, repetitions, and false starts.
3. Check proper names, numbers, and technical terms, which are exactly where voice recognition makes mistakes.

The operational trick is not to rewrite by hand: **re-slice**. If you keep the word-by-word timestamps from the transcription, changing the maximum characters or the CPS and regenerating the file takes seconds; correcting cue by cue takes an afternoon.

## What changes (and what doesn’t) with CapCut AI in November 2026

Current events reinforce the argument rather than dismantle it: AI has automated transcription, not judgment. On November 11, 2026, the [version 20.1.0 of CapCut for Android](https://apkmirror.com/apk/bytedance-pte-ltd/capcut/capcut-video-editor-20-1-0-release) was released, and CapCut's own page about the [latest APK version](https://www.capcut.com/resource/capcut-apk-new-version) describes a voice-to-text tool with AI that "automatically transcribes speech into accurate subtitles".

On the same day, at Mexico Tech Week 2026, ByteDance presented CapCut PC with **Omni Agent**, a desktop agent that coordinates complete production flows from an idea, with sub-agents like Video Studio and Design Studio and an EditPilot that converts instructions in natural language into multi-layer edits on the timeline, according to the [event report published by La Crónica de Hoy](https://cronica.com.mx/tecnologia/2026/11/11/capcut-pc-como-puedes-convertir-una-idea-en-un-proyecto-completo-con-ia) on November 11, 2026. The presentation itself framed AI as an "scaling engine" and human judgment as "strategic control".

It is exactly the division of tasks of this article. The machine gives you the words and timestamps. The maximum characters, the number of lines, the CPS, the minimum duration, and the gap are decided by you, as they depend on the audience, the language, and the framing. In short: **transcription is already free; slicing remains yours**.

## Subtitles are now a legal obligation, not just a good practice

As of **June 28, 2025**, the [Directive (EU) 2019/882](https://eur-lex.europa.eu/eli/dir/2019/882/oj), the European Accessibility Act, applies, extending accessibility requirements to consumer digital products and services. In Spain, it is transposed by the [Law 11/2023, of May 8](https://www.boe.es/buscar/act.php?id=BOE-A-2023-11022), published in the BOE, whose obligations are enforceable from that same date; microenterprises are exempt from the duty to comply.

The technical reference framework for demonstrating compliance is the harmonized standard **EN 301 549**, which translates "accessible" into verifiable requirements.

This is supplemented by the scope argument, which has been documented for years: in a study by Verizon Media and Publicis Media with 5,616 American adults aged 18 to 54, conducted in April 2019 and [reported by Forbes](https://www.forbes.com/sites/tjmccue/2019/07/31/verizon-media-says-69-percent-of-consumers-watching-video-with-sound-off/) on July 31, 2019, **69%** reported watching video without sound in public spaces and **80%** said they were more likely to finish a video if it had subtitles. These are 2019 data from the U.S., and it is advisable to read them as an order of magnitude, not as a measurement of your audience.

The consequence is twofold: an illegible subtitle not only annoys, but also violates the spirit of the standard. And legibility is not declared: it is measured with the parameters of this article.

## There is no official standard for vertical video subtitling

As of today, there is no specific subtitling standard for 9:16 comparable to broadcast guidelines. The closest is the adaptation of existing standards made by the BBC in its version 1.2.5 and the work of the localization sector.

As early as November 2019, Belén Agulló García pointed out in [an analysis published by Nimdzi](https://www.nimdzi.com/subtitling-vertical-videos-guidelines-where-art-thou/) that "there is still no standard for characters per line in vertical videos" and proposed starting from the BBC's limit, maintaining one or two lines at most, establishing a safe zone so that the text is not cut off on different screens, and editing the text —removing repetitions and filler words— instead of transcribing literally when space is tight.

Seven years later, the diagnosis remains valid, with one difference: the BBC now incorporates explicit guidance for 9:16 (size, position, and number of lines). Until there is more, the reasonable practice is summarized in this table:

| Parameter | Working value in 9:16 | Origin |
|---|---|---|
| Characters per line | 28-34 (18-22 in a single line) | BBC equivalence of 37 → 25 characters, adjusted to social media font |
| Lines per cue | 2 as a ceiling | BBC allows 3 in 9:16; reduced to 2 due to interface and face |
| Reading speed | 12-17 CPS | Below Netflix's 20 CPS for adults |
| Minimum duration | 0.8-1 s | BBC ~0.3 s/word; Netflix 20 frames |
| Maximum duration | 4-6 s | Avoid re-reading as warned by the BBC |
| Gap between cues | 0.05-0.1 s chained; ≥1 s if it is a real pause | Netflix 2 frames; BBC 1 s minimum on pause |
| Position | Lower third, raised from the edge | BBC, section on vertical position |
| Line height | 3.9-4.5 % of the video height | BBC, authoring font size |

## Frequently Asked Questions

### How many characters per line should a subtitle in vertical video have?

Between 28 and 34 characters works well in 9:16. The reference is the BBC equivalence: 37 characters in a 16:9 correspond to about 25 in a 9:16 at 90% width. If you use large single-line subtitles, drop to 18-22.

### What are CPS in subtitles and what is the correct value?

CPS stands for characters per second that the viewer has to read: cue characters divided by their duration. Netflix allows up to 20 CPS for adult content and 17 for children; for vertical social video, 12-17 CPS is the comfortable range, with 15 as a starting point.

### How long should a subtitle last at a minimum?

Netflix requires a minimum of 20 frames, which is 4/5 of a second. The BBC suggests it by words: about 0.3 seconds per word, 1.2 seconds for a subtitle of four words. In vertical, 0.8-1 second is a sensible minimum.

### Can subtitles have three lines in vertical video?

The BBC allows it: recommends a maximum of two lines in 16:9, 4:3, and 1:1, and three in 9:16. In social video, however, the third line usually covers the face or clashes with the application interface, so the practical limit is two lines.

### What gap should be left between two subtitles?

Two different rules. For chained cues, Netflix asks for a minimum of two frames and prohibits intermediate gaps of 3 to 11 frames; in practice, 50-100 milliseconds constant. If there is a real pause in the speech, the BBC asks for at least one second, preferably one and a half seconds.

### Why is my .srt not importing in the editor?

It is almost always the file format: milliseconds with a point instead of a comma, numbering with jumps, overlapping cues, or inconsistent line breaks. Validate the numbering from 1, the decimal separator (comma), the increasing times without overlaps, and the blank line between cues.

### Are the automatic subtitles from the editing app sufficient?

For the words, usually yes. For the chunking, no: voice recognition cuts by audio pauses, not by box width or reading speed. The efficient thing is to keep the transcription and chunk it again with your own parameters.

### Are subtitles legally required?

For digital products and services included in its scope, yes from June 28, 2025, under the European Accessibility Act (Directive (EU) 2019/882) and, in Spain, under Law 11/2023. Micro-enterprises are exempt from this obligation. The technical compliance criterion is the EN 301 549 standard.

## Set the rules before importing, not after

The subtitle that is read is not the one with the best words: it is the one with the correct chunking. Set your values —maximum characters, two lines, CPS, minimum and maximum duration, gap— before generating the file, and review the `.srt` as a data point, not as text.

That is the part that **aiCapCut**, from **AiToolsForApps**, converts into a form: you paste your script or upload the audio, choose the chunking parameters, and get a validated `.srt`, along with its `.vtt` and `.txt` variants, ready to import. It brings presets designed for vertical format and allows **re-chunking the same material with different values without regenerating anything**, which is exactly what saves the afternoon of corrections. It does not render video or touch your timeline: it gives you the file. You can try the [SRT subtitle tool and editing plan for vertical video](/capcut-srt-subtitles-edit-plan-png-titles-245) without registering. aiCapCut is an independent tool and is not affiliated with CapCut or ByteDance.

> Generate your .srt with your own segmentation rules — characters per line, CPS, minimum duration, and gaps — and download it validated in .srt, .vtt, and .txt. No registration required to start.

- [CapCut — .srt subtitles without typing, shot-by-shot edit plan, PNG titles and export dimensions](https://aitoolsforapps.com/capcut-srt-subtitles-edit-plan-png-titles-245)

## About the platform

**CapCut — .srt subtitles without typing, shot-by-shot edit plan, PNG titles and export dimensions** — an app by AiToolsForApps. Available at: https://aitoolsforapps.com/capcut-srt-subtitles-edit-plan-png-titles-245/blog/vertical-video-subtitles-characters-per-line-cps

### Other languages

- `es`: https://aitoolsforapps.com/es/capcut-subtitulos-srt-plan-de-montaje-y-rotulos-245/blog/subtitulos-video-vertical-caracteres-por-linea-cps
- `fr`: https://aitoolsforapps.com/fr/capcut-sous-titres-srt-plan-de-montage-et-titres-png-245/blog/sous-titres-video-vertical-caracteres-par-ligne-cps
- `it`: https://aitoolsforapps.com/it/capcut-sottotitoli-srt-piano-di-montaggio-e-titoli-png-245/blog/sottotitoli-video-verticale-caratteri-per-riga-cps
- `pt`: https://aitoolsforapps.com/pt/capcut-legendas-srt-plano-de-montagem-e-rotulos-png-245/blog/legendas-video-vertical-caracteres-por-linha-cps
- `de`: https://aitoolsforapps.com/de/capcut-srt-untertitel-schnittplan-und-png-titel-245/blog/untertitel-vertikale-videos-zeichen-pro-zeile-cps
- `bg`: https://aitoolsforapps.com/bg/subtitri-srt-za-capcut-bezplatno-plan-za-montazh-i-png-nadpisi-245/blog/subtitri-vertikalno-video-znatsi-na-red-cps
- `sv`: https://aitoolsforapps.com/sv/srt-undertexter-for-capcut-gratis-klippplan-och-png-rubriker-245/blog/undertexter-staende-video-tecken-per-rad-cps
- `ja`: https://aitoolsforapps.com/ja/capcut-srt-jimaku-shot-list-png-245/blog/tate-douga-jimaku-1gyo-mojisu-cps
- `nl`: https://aitoolsforapps.com/nl/srt-ondertitels-voor-capcut-gratis-montageplan-en-png-titels-245/blog/ondertitels-verticale-video-tekens-per-regel-cps
- `ko`: https://aitoolsforapps.com/ko/capcut-srt-subtitle-edit-plan-png-lower-third-245/blog/sero-yeongsang-jamak-julttang-geulja-su-cps
- `da`: https://aitoolsforapps.com/da/srt-undertekster-til-capcut-gratis-klipningsplan-og-png-skilte-245/blog/undertekster-lodret-video-tegn-pr-linje-cps
- `pl`: https://aitoolsforapps.com/pl/napisy-srt-do-capcut-za-darmo-plan-montazu-i-napisy-png-245/blog/napisy-wideo-pionowe-znaki-w-linii-cps
- `fi`: https://aitoolsforapps.com/fi/ilmaiset-srt-tekstitykset-capcutiin-leikkaussuunnitelma-ja-png-tekstit-245/blog/pystyvideon-tekstitys-merkkeja-rivilla-cps
- `id`: https://aitoolsforapps.com/id/subtitle-srt-untuk-capcut-gratis-rencana-editing-dan-teks-overlay-png-245/blog/subtitel-video-vertikal-karakter-per-baris-cps
- `he`: https://aitoolsforapps.com/he/ktobiot-srt-l-capcut-bchinm-toknit-arikh-vkitobi-png-245/blog/ktuviyot-sirton-anechi-tavim-lesura-cps
- `zh`: https://aitoolsforapps.com/zh/capcut-srt-subtitles-edit-plan-overlays-245/blog/zhimu-zhishi-yingpian-meihang-ziyuanshu-cps
- `no`: https://aitoolsforapps.com/no/srt-undertekster-for-capcut-gratis-klippeplan-og-png-skilt-245/blog/undertekster-staende-video-tegn-per-linje-cps
- `ar`: https://aitoolsforapps.com/ar/trjmat-srt-l-capcut-mjana-wkhta-muntaj-wlaftat-png-245/blog/tarjamat-fidyu-amudi-ahruf-fi-assatr-cps
- `cs`: https://aitoolsforapps.com/cs/titulky-srt-pro-capcut-zdarma-plan-strihu-a-png-napisy-245/blog/titulky-vertikalni-video-znaky-na-radek-cps
