# Slopify 3.0.4

Released 28 September 2026.

3.0.4 keeps long videos moving: subtitles stay in sync through hard stretches instead of stopping the video, Slopify no longer freezes during a long run, and audiobooks and podcasts sound even from one voice to the next. Play's settings are laid out in one tidy grid.

## Highlights

- **Subtitles stay in sync** through years, names and abbreviations the voice says its own way.
- **No more freezes** while a long video is being made.
- **Even voices** in audiobooks and podcasts: every line within 1 LU of the narrator.
- **Visible buttons** to make a stage again, and **a waveform** in the audio player.

## Subtitles

- Years from 1000 on are read as a narrator says them ("eleven fifty seven", "eleven o four", "thirteen hundred"), so history and fantasy dates match their narration.
- A short stretch the voice says differently from the text, such as a name or an abbreviation, is placed between the words around it rather than stopping the video.
- When timing does stop, the message says where to listen, and to regenerate the chunk only if words are missing; a voice that read it right, just another way, reads it the same way again.

## Speed

- Slopify no longer stops answering for seconds at a time while a long video is made. The article used to be read again from scratch several times a step.

## Audiobooks and podcasts

- Every speaker's line is levelled to within 1 LU of the narrator's in the audio export, the MP3 and M4B and the video, so no character sits louder than the narrator. The gain changes in the pause between lines and the captions keep their timing.
- With several speakers the single voice at the top of Narration is hidden when there is no intro or outro, and labelled **Intro and outro voice** when there is.
- When a provided book only needs a text model to work out who speaks each line, **Text generation** shows under Narration and says what it is used for.
- **Script by speaker** is in the Download menus of Narration and Article.

## Project page

- Making a whole stage again is a button on its section: **Export the audio again**, **Render the video again**, **Render the PDF again** and the rest. Each still asks first.
- The Narration stage says when it plays a levelled narration an edit made outdated, and plays the new one beside it until the run levels it.
- A paused run lists what it still has to make.
- When the Library's intro or outro has changed since a project copied it, its picker in **Edit settings** lists both: **this project's version** and the **Library version (newer)**. Prompts under **Prompts** say so too, with **Use the Library version** or **Keep this project's version**. Intros and outros were missed before.
- The audio player shows the narration's waveform.

## Play

- Every section's settings sit in one grid: labels above, equal columns, three or four across a wide screen. Every dropdown and number box is the same size, and image prompts line up in columns.

## Fixes

- The PDF's Sources page reads as text; only the web addresses are links, and an entry that cites several addresses links every one.
- Pauses between sentences weigh a number by how long it takes to say, skip a name's initial and never lengthen the short breath between two words.

## Upgrading from 3.0.3

- **Nothing becomes outdated.** Audiobooks and podcasts made before get the even voices when their audio is exported again (**Export the audio again**).