Versely
    Guides

    The French voice-over page ships six named voices, not a French presenter

    /voice-over/french exists because six MiniMax names clear the roster floor (English ships 80). It is a voice list plus engines. It is not a generate of a talking French host.

    Versely Team••5 min read

    /voice-over/french exists because six MiniMax names clear the roster floor (English ships 80). It is a voice list plus engines. It is not a generate of a talking French host.

    The language page is a roster surface. It lists who can speak, on which engines, in French. It does not cast a face. It does not open Veo. Asking a video model for "a French presenter in a studio" is a different catalog row and a likeness you probably do not have.

    Six named MiniMax voices is the page

    The gate is arithmetic. A language page ships when the MiniMax roster is named — not generic Male Voice / Female Voice placeholders — and enough engines actually speak the language. French clears with six names:

    • Male Speech
    • Female Speech
    • Casual Man
    • Movie Lead Female
    • Female Anchor
    • Male Narrator

    Those six strings are the unique media. English contributes eighty named voices. Spanish twenty. Korean fifteen. French contributes six. Below a handful of names there is not enough page left to be distinct, which is why German's three names did not ship as a twin of this one. The voice-over index is the map of the four language pages that did: English, Spanish, Korean, French.

    This blog is not a second French page. It does not reprint the engine table. It answers a different query: why the URL exists, and why it is not a talking-host generate.

    Pick a name. Audition the same line in two of them. Keep the one that hits the consonant. Do not pick "French" as a vibe and let a video model invent a person who happens to speak it.

    Seven engines, a speech file, not a mouth

    French is covered by all seven speech engines in the app: MiniMax, ElevenLabs, Cartesia, Inworld, Gemini, Qwen, Grok. That is engine count, not a presenter count. Seven ways to render a script is still one job: audio.

    MiniMax Speech is the named-roster row: five, ten, thirty, or sixty seconds, listed at 2 credits, discounted. Cast a specific voice. It is not Hailuo. Hailuo is MiniMax's video line. This row is speech.

    Codes differ by engine — fr, FR_FR, fr-FR, French — and they belong on the language page, not in a Veo prompt. French is also a dubbing target. Dubbing is a later job on a picture you already locked. It is not a reason to generate a host.

    Cloning paths exist on some engines from a sample or a written description. Cloning a voice is still a speech file. It is not a face. Do not treat a cloned French read as permission to lipsync a constructed presenter.

    Gemini Flash TTS is a speech row

    Gemini 3.1 Flash TTS is text-to-speech: 4 credits, thirty voices on the engine, natural-language style control, inline tags ([sigh], [laughing], [whispering]), multilingual synthesis, multi-speaker dialogue. That is a read. It is not native audio on a video generate, and it is not thirty French presenters.

    Casting French here is a language flag plus a voice. Style instructions steer delivery — warm, fast, whispered — without opening a talking-head model. Multi-speaker dialogue is two reads in one file, still audio. If you needed a face, you needed a different row and a grant.

    Do not prompt Gemini TTS to "be a French news anchor on camera." The model does not have a camera. Female Anchor on the MiniMax list is a voice name, not a person you can film.

    The agent writes audio, not a face

    Write and generate a voiceover is the named agent job: give it a script, a mood, a language, a voice_id if you have one. It generates a spoken track. Cost is priced per voice model, shown before you confirm. The output is an audio clip, ready to attach to a picture you already have or to a lipsync you separately authorised.

    That is the whole product. It is not "make me a French host." It is "speak these words." Attach the file to a locked plate — a product still, a captioned explainer, b-roll you filmed. If the picture still needs a mouth, that mouth is a real named person or a likeness you cleared. The French voice-over page does not clear a face.

    Do not:

    • Open a talking-head video row and prompt "professional French presenter."
    • Treat the six MiniMax names as characters you can draw.
    • Twin this language as a second /voice-over/french with a different title.

    The roster is the point. Six names. Seven engines. A file that speaks. No host.

    FAQ

    Why does French ship when Italian has seven named voices?

    Roster depth is not the only gate. Pages that measured as near-duplicates of a deeper language did not ship. French cleared both the named-voice floor and the twin test. Italian did not. This URL is not a ranking of how well the app speaks French. It is a page the roster could actually fill.

    Can I use Gemini 3.1 Flash TTS as my French presenter on camera?

    No. Gemini 3.1 Flash TTS is a speech row: 4 credits, thirty engine voices, multilingual reads, inline tags. It returns audio. A talking French host is a video generate plus a likeness. Cast a voice from the French page. Film a person, or do not put a face on the line.

    Is MiniMax Speech the same as Hailuo?

    No. MiniMax Speech is the TTS roster — 2 credits listed, discounted, 5s / 10s / 30s / 60s. Hailuo is MiniMax's video family. Do not pick Hailuo because a French voice demo felt warm.

    Does the agent generate a talking French host if I ask for "a presenter"?

    The named job is script-to-speech. It writes a voiceover file. It does not mint a person. If you needed a face, that is a different capability and a grant. Stay on the roster.