Narration is a feature that many customers have requested for a long time. "Narration cuts off every time I pause for a shape," "Subtitles are short and turn into fast speech that's hard to understand," "Volume changes from sentence to sentence" — all of these have been requested repeatedly, but they couldn't be fixed without redesigning how the audio itself is created.

This time, we've redesigned the entire process from creating audio to playing it back, and we're addressing all three of these issues at once. We're also adding a workflow for creating subtitles from step descriptions and a way to identify fast-speech subtitles in the editing screen. Thank you for your patience.

① Subtitle narration no longer stops even while paused

When you add shapes (arrows, circles, spotlights, and so on) to a video, the video pauses at that position and displays the shape with a "Still Duration." Previously, narration would also pause during this still time, causing sentences being read to be cut off at that point, with the rest continuing after the pause. If there are multiple pauses within a single subtitle, the text would be fragmented each time.

Now, narration continues reading as-is even while the video is paused. Sentences no longer get cut off in the middle.

● Reading order does not change

• Reading continues through the range the subtitle covers (until just before the next subtitle begins). The next subtitle is not read early
• After finishing reading, it waits silently for the video to catch up
• For steps without subtitles or if the video pauses between subtitles, it remains silent as before

② Fast-paced subtitles are now easier to understand

Subtitle narration speed is adjusted to fit within the length of time the subtitle is displayed. When text is long and the frame is short, the created audio was previously sped up after creation to fit. Additionally, the playback side also stretches and compresses it, resulting in double time manipulation that caused the audio to sound distorted like an echo.

This time, we've changed the compression method in three ways.

● We read faster from the start rather than compressing afterward

For subtitles that won't fit in the frame, we now read them slightly faster when creating the audio. With less post-processing speedup needed, the audio sounds clearer.

● We use the gap to the next subtitle

When there's a gap between a subtitle and the next one, we now read through to that point. For example, if a 1.9-second frame contains 6 seconds of text, we previously compressed it by more than 3 times, but by using the gap, we only need to compress by about half. Since we don't exceed the start position of the next subtitle, the subtitle and audio timing stays in sync.

● We use shape pauses to restore the reading speed

We use the ability to read through pauses from ① to restore the fast speech. When there's a shape pause in the step, compressed subtitles are read more slowly by that pause duration. The still time for shapes is now both a time to show caution and a time to create breathing room for narration.

③ Narration volume is now consistent

Subtitle narration is created sentence by sentence and then connected, which resulted in volume differences between sentences. Even within the same SOP, there would be drops where "this part is too quiet to hear" and jumps where "the next sentence suddenly gets loud," and volume would change every time you switched between SOPs or languages.

Now, we normalize the volume of each sentence before arranging them. We're following the volume standards used in broadcasting, so volume remains nearly the same across different SOPs and when switching to translated languages. We've set an upper limit on how much we increase the volume to prevent distortion.

④ You can create subtitles from step descriptions

For SOPs where you've already written step descriptions, you can now create subtitles from that text as-is. In the subtitle list in the video editor, select [From the step text] from [Create subtitles]. For a single step, you can also create them from the button in the description field.

● Nothing gets crammed into one frame, so fast speech doesn't happen

Rather than putting an entire description into a single subtitle, sentences are split and distributed across the step's duration. Since long text isn't crammed into a single frame, the fast-speech problem from ② is less likely to occur in the first place.

You can also work in reverse. You can bulk transfer subtitle and shape text to the step description, so you can start from either the video side or the text side and keep them in sync.

⑤ The editing screen shows which subtitles are fast-paced

Previously, you didn't know which subtitles were fast-paced until you played them back and listened. Now you can check directly in the subtitle list in the video editor.

• Subtitles where narration doesn't fit in the frame get a "fast ○x" marker
• The number is displayed after speed adjustment in ②. Subtitles that are actually read slowly because there's a pause are not counted as fast
• For steps with no fast subtitles, that's displayed instead

For marked subtitles, widening the frame or splitting the text into shorter segments will result in natural pacing.

About applying changes to existing SOPs

The feature from ① where "narration doesn't stop even during pauses" is already enabled for existing SOPs as-is. No recreation is needed.

The speed from ② and volume from ③ are applied when narration audio is recreated. Audio is recreated when you rewrite subtitles and save, change the original language, register in the Pronunciation Dictionary, or perform Bulk replace. Until it's recreated, the previous audio plays back as-is.

Thank you for using Dive ◎