Sentioscope

Speech and voice controls, as each vendor documents them

Ranking of tools by the voice controls they document

One ranking sorts the register's 16 tools by how many of five voice controls their documentation states. HeyGen, PixVerse and SceneMixer share the lead. As of 2026-09-12.

From a reference page to a placeA tool's reference is checked against five voice questions, each answered one earns a point, and the tools are sorted by the total.Five questionsAbout the voicein a scene.ItemsA referencepageChecked foreach, silencemarked.A tallyA sorted tableMost controlsfirst, tiessharing aplace.PlacesThe pagesquotedWording behindthe leadingplaces.
Fig. 1 The table is a sorted tally of documented controls, with equal totals sharing a place.

The ranking checks every tool's reference against the same five voice questions and gives one point for each answered. Where the reference is quiet the register says so, and the table scores the silence as nothing.

A documented control is not a good control. A tool that explains its voice options at length can outscore one with excellent voices and a short page, so the ranking reads as a map of what is written down. Its own page lists the five controls, how ties show and what the count cannot say.

Several tools can sit on one total; they share a place and appear in alphabetical order, which means nothing. The pages behind the leading places are quoted beneath the table with their addresses and the day each was read.

A tool is listed only when its reference documents at least one control, and each row links to the tool's page in this register, where every reading is quoted in full. Nothing in the table is estimated from a demo, a blog post or a price list: a control the reference does not describe is recorded as undocumented and earns no point.

1The rankings

  • Character voices — Five voice controls, from where the audio comes from to lip-sync, counted by what each reference documents.

Rows are tools whose references were read; the reading notes explain what counts as documented, and the models hold every reading.