Who publishes a claim about text drawn in the picture
One vendor in this register publishes a claim about text drawn inside the picture, naming English and Chinese before saying other languages. No other vendor here addresses it, and it is the one language layer nothing can repair afterwards. As of 2026-09-12.
| Tool | What is published | Detail |
|---|---|---|
| PixVerse | Multilingual text generation within frames, with accurate placement | Two languages named, then others |
| PixVerse | No voice list published | The voice cell is empty |
| Everyone else | No claim about text inside the picture | The layer is undocumented across the register |
Inclusion rule. Vendors publishing any claim about text rendered inside the frame, plus the rest of the field as one row. Subtitle claims are a different layer and excluded. Order. The vendor publishing a claim first.
1It is the layer with no repair available at all
A sign, a note or a phone screen generated in one language stays in that language. Subtitles sit on top of the picture and dubbing does not touch it, so changing it means generating the shot again.
That makes it the most consequential language decision in a generated pipeline and the least documented. One claim exists in this register and it names two languages before saying others.
2Scripts are where the capability breaks
Drawing legible characters is about glyph shapes, spacing and joining. A model trained largely on Latin and Chinese characters is not evidence about a diacritic-heavy script, an abugida or one that runs right to left.
The claim names one Latin-script language and one Chinese-script language. It cannot be read as covering the harder cases or as excluding them, and the register records it as published rather than interpreting it.
3Why it is not a column
A column needs enough vendors publishing comparable material that an empty cell means something. One claim from one vendor would make a column where seventeen blanks implied a question had been asked of everybody.
So it is recorded beside the empty voice cell on that vendor page, and the practical advice follows from the layers rather than from the claim: keep legible text out of generated frames wherever the story does not need it.
4The three columns behind this question
| Column | How many of the 18 tools name a list |
|---|---|
| Dialogue language | 5 of 18 tools |
| Document language | 1 of 18 tools |
| Site language | 9 of 18 tools |
Inclusion rule. The three columns this register keeps. A column stays listed even when no vendor names a list, because silence in a column is itself recorded. Order. Alphabetical by column name.
- Text drawn inside the picture“Multilingual text generation within frames is now supported, with accurate placement and style consistency across English, Chinese, and other languages”outside the three columns
- Voice languagesNo public list (as of 2026-09-12)no list published
- Site languages13 languages: de, en, es, fr, id, ja, ko, pt, ru, th, tr, vi, zhdeclared as same-domain alternates
5Sources
Counts and lists are read from the vendor pages named in the register, checked 2026-09-12. The counting rule is on how counted. Nearby: A code that points elsewhere, Sets inside sets.