Hailuo on where the sound comes from
Hailuo holds one of two fives for sound. A five needs speech generated with the picture, in named languages, with voice selection, and naming the languages is the part most entries here do not do. As of 2026-09-12.
| Point | As the rubric has it, or as the table records it |
|---|---|
| The question this column asks | Where does the audio come from? |
| What this tool rates | 5 of 5 |
| Tools holding the same value | Two of the sixteen rated here |
| What a five looks like | Speech generated with the picture, in named languages, with voice selection |
| What sits at the bottom of the scale | Silent output with no statement about audio at all |
| Tools left unrated on this column | None of the sixteen |
Inclusion rule. The rubric's own wording for this criterion, with this tool's value beside it. Nothing here is taken from a demo, a benchmark or a review. Order. The question first, then the value, then the rule it was given under.
1The whole column, with this row marked
| Tool | Sound |
|---|---|
| CapCut | 4 of 5 |
| Dreamina | 3 of 5 |
| Hailuo (MiniMax) | 5 of 5 |
| invideo AI | 4 of 5 |
| Kling AI | 3 of 5 |
| LTX Studio | 3 of 5 |
| Luma Dream Machine | 3 of 5 |
| Morph Studio | 2 of 5 |
| Pika | 2 of 5 |
| PixVerse | 3 of 5 |
| Runway | 3 of 5 |
| SceneMixer | 4 of 5 |
| Sora (OpenAI) | 4 of 5 |
| Veo / Flow (Google) | 5 of 5 |
| Vidu | 2 of 5 |
| Wan (Tongyi Wanxiang) | 4 of 5 |
Inclusion rule. Tools rated on this rubric: a public English product page addressing at least four of the seven criteria. An unrated cell means nothing usable is published on that point. Order. Alphabetical by tool name.
2Naming the languages is what separates a five from a four
Five tools sit at four in this column, which is generated audio without the specifics. A multilingual claim with no list cannot be planned against: a production shipping to a particular market needs to know whether its language is in scope before committing, and a claim does not answer that.
So the five is a documentation value, like everything on this rubric. It says the information a localisation plan needs is published, not that the voices are better.
3Sound arriving with the picture removes a whole pipeline stage
Where speech is generated with the shot, lip movement and audio come from the same process and there is nothing to align afterwards. Where they are separate, a production adds a voice step, a sync step and a review step, and the sync step is where AI drama most visibly fails.
That is why this column is worth reading separately from picture quality. A tool with a slightly weaker image and native speech can be cheaper per finished minute than a stronger one that leaves audio outside the rate.
4The rest of this row is the widest spread in the table
Fives on picture and sound, ones on script handling and assembly. Strong exactly where generation happens and silent on both sides of it, which for forty episodes means the two ends of the work stay with a person and appear on no rate card.
For a single strong shot that shape is ideal. The five in this column is the reason to put this row on a shortlist and the ones either side of it are the reason the shortlist needs a second tool.
Ratings describe what each vendor publishes on its own pages, read 2026-09-12, and are given under the rule on how scoring works. This column across every tool is on sound; this tool across all seven is on Hailuo (MiniMax). Nearby cells: Veo and Flow, sound, LTX Studio, assembly. All of them: single ratings.