Speech and Dialogue Move to Eleven v4 — Same Media Band

October 9, 2026

Spoken replies, chat voice-overs, and multi-speaker Dialogue now come from Eleven v4, ElevenLabs' newest expressive speech model. You ask the same way you always have. A call still costs the same.

What you can do

• Hear Eleven v4 by default. Every Speech call, every voice-over the chat plays, and every Dialogue recording is generated on Eleven v4. You do not name the model.
• Keep the same voices and delivery. Alfred and Ada are unchanged. The six styles are still excited, thoughtful, confident, analytical, gentle, dramatic. Inline tags such as [excited], [whispers], [laughs], and [sighs] still shape a line.
• Write the line in the language you want spoken. Eleven v4 covers 90+ languages, including Hindi. Put the words in that language. There is no separate language setting on Speech.
• Get a backup without asking. If Eleven v4 cannot take the call, Alfrada tries Eleven v4 Turbo, the real-time model in the same family, and then the older multilingual model. You do not pick Turbo. A normal reply is v4.
• Reuse the MP3. The file still lands in the session, and the Video Editor can still lay it under a picture.

Where this shows up

You ask for a spoken summary of a brief. You did nothing differently, and the voice-over that plays is from a newer speech model than the one you heard last month.

You want Ada to narrate a deck, a few sentences per slide, thoughtful and confident. The clips are still short, still Ada, and still files you can drop into the Video Editor.

You write a line in Hindi and ask Speech to say it. The same Alfred or Ada voice speaks those words. You do not switch models to get the language.

Try it

• "Use the Speech tool. Say this out loud, excited: I found something incredible, there is a forty percent spike."
• "Narrate this paragraph in Ada's voice, thoughtful, with a pause after the second sentence."
• "Use Speech. Say this in Hindi: नमस्ते। आज का नंबर बारह प्रतिशत ऊपर है।"

Heads up

• Speech and Dialogue stay on Pro and above, in the Media band at 500,000 tokens per call. Moving to Eleven v4 does not change the band.
• Dialogue is still the multi-speaker tool. One voice is Speech. A script with more than one speaker is Dialogue. Both use Eleven v4. A long Dialogue script is still joined into one MP3, and v4 keeps the voices continuous across those joins.
• A clip is still short — about 150 words. A long script is several Speech calls, or one call per slide, not one ten-minute file.
• Voice-overs you already generated keep playing. Only new calls use Eleven v4.