Simba 3.5

48 languages.
Spoken like a local.

or scroll to meet them

One model idEvery voice speaks all 48

The thread

None of these voices are human.

Each line below is a single request to simba-3.5, spoken by a voice native to its language. Follow the thread.

Et voilà,

toute chaude !

Bonne journée !

Paris

A baker, handing over a baguette

“There you go, still warm! Have a nice day!”

Alexandre fr-FR · simba-3.5

View request

POST https://api.speechify.ai/v1/audio/speech

json

{
  "model": "simba-3.5",
  "voice_id": "alexandre",
  "language": "fr-FR",
  "input": "Et voilà, toute chaude ! Bonne journée !",
  "audio_format": "mp3"
}

いらっしゃいませ！

何名様ですか？

Tokyo東京

An izakaya owner, greeting guests at the door

“Welcome! How many of you?”

Daiki ja-JP · simba-3.5

View request

POST https://api.speechify.ai/v1/audio/speech

json

{
  "model": "simba-3.5",
  "voice_id": "daiki",
  "language": "ja-JP",
  "input": "いらっしゃいませ！何名様ですか？",
  "audio_format": "mp3"
}

Шапку надень!

На улице

минус двадцать.

MoscowМосква

A mother at the door in winter, holding out a hat

“Put your hat on! It’s minus twenty out.”

Darina ru-RU · simba-3.5

View request

POST https://api.speechify.ai/v1/audio/speech

json

{
  "model": "simba-3.5",
  "voice_id": "darina",
  "language": "ru-RU",
  "input": "Шапку надень! На улице минус двадцать.",
  "audio_format": "mp3"
}

밥은
먹었니?

Seoul서울

A grandmother, opening her door

“Have you eaten?”

Bo-kyung ko-KR · simba-3.5

View request

POST https://api.speechify.ai/v1/audio/speech

json

{
  "model": "simba-3.5",
  "voice_id": "bo-kyung",
  "language": "ko-KR",
  "input": "밥은 먹었니?",
  "audio_format": "mp3"
}

გაუმარჯოს
ჩვენს
სტუმრებს!

Tbilisiთბილისი

The host of a supra, raising a toast

“To our guests!”

Lasha ka-GE · simba-3.5

View request

POST https://api.speechify.ai/v1/audio/speech

json

{
  "model": "simba-3.5",
  "voice_id": "lasha",
  "language": "ka-GE",
  "input": "გაუმარჯოს ჩვენს სტუმრებს!",
  "audio_format": "mp3"
}

48 languages

Drag across all 48 languages. Every voice you hear is native.

Native voice

Quiet here: no native voice yet

For the best results, use a voice native to the language. A voice keeps the accent of the recording it was made from.

Now playing العربية · Arabic

مرحبًا، شكرًا لاتصالك. يمكنني تغيير حجز، أو متابعة طلب، أو الإجابة عن سؤال حول حسابك.

“Hi, thanks for calling. I can change a booking, check on an order, or answer a question about your account.”

Aseel ar-AE · simba-3.5

Languages without a native voice yet stay quiet here. Any voice reads them through the API.

Native voices

Simba 3.5 sounds best with a voice native to the language you generate. Every voice keeps the accent of the recording it was made from, so an English voice reading German sounds like an English speaker speaking German. A cloned voice works the same way: clone from a recording in the language you will use it in.

We kept this on purpose. Customers asked that a voice, and above all their own clone, keeps sounding like itself in every language. We plan to make this more flexible in an upcoming release.

For each language, start with its top voices.

Build it

One request. Change the voice.

Same endpoint, same model. Swap voice_id, language and input, and each voice reads its own language. The response carries speech marks: the timings behind every highlighted word on this page.

Get an API key Read the docs

POSThttps://api.speechify.ai/v1/audio/speech

Authorization: Bearer $SPEECHIFY_API_KEY

{
  "model": "simba-3.5",
  "voice_id": "alexandre",
  "language": "fr-FR",
  "input": "Et voilà, toute chaude ! Bonne journée !",
  "audio_format": "mp3"
}

200response

{
  "audio_data": "SUQzBAAAAAAAI1RTU0UAAAAPAAADTGF2…",
  "audio_format": "mp3",
  "speech_marks": { "chunks": [
    { "value": "chaude !", "start_time": 840, "end_time": 1523 },
    …
  ] }
}

POSThttps://api.speechify.ai/v1/audio/speech

Authorization: Bearer $SPEECHIFY_API_KEY

{
  "model": "simba-3.5",
  "voice_id": "daiki",
  "language": "ja-JP",
  "input": "いらっしゃいませ！何名様ですか？",
  "audio_format": "mp3"
}

200response

{
  "audio_data": "SUQzBAAAAAAAI1RTU0UAAAAPAAADTGF2…",
  "audio_format": "mp3",
  "speech_marks": { "chunks": [
    { "value": "し", "start_time": 456, "end_time": 499 },
    …
  ] }
}

POSThttps://api.speechify.ai/v1/audio/speech

Authorization: Bearer $SPEECHIFY_API_KEY

{
  "model": "simba-3.5",
  "voice_id": "darina",
  "language": "ru-RU",
  "input": "Шапку надень! На улице минус двадцать.",
  "audio_format": "mp3"
}

200response

{
  "audio_data": "SUQzBAAAAAAAI1RTU0UAAAAPAAADTGF2…",
  "audio_format": "mp3",
  "speech_marks": { "chunks": [
    { "value": "улице", "start_time": 1437, "end_time": 1864 },
    …
  ] }
}

POSThttps://api.speechify.ai/v1/audio/speech

Authorization: Bearer $SPEECHIFY_API_KEY

{
  "model": "simba-3.5",
  "voice_id": "bo-kyung",
  "language": "ko-KR",
  "input": "밥은 먹었니?",
  "audio_format": "mp3"
}

200response

{
  "audio_data": "SUQzBAAAAAAAI1RTU0UAAAAPAAADTGF2…",
  "audio_format": "mp3",
  "speech_marks": { "chunks": [
    { "value": "먹었니?", "start_time": 499, "end_time": 1736 },
    …
  ] }
}

POSThttps://api.speechify.ai/v1/audio/speech

Authorization: Bearer $SPEECHIFY_API_KEY

{
  "model": "simba-3.5",
  "voice_id": "lasha",
  "language": "ka-GE",
  "input": "გაუმარჯოს ჩვენს სტუმრებს!",
  "audio_format": "mp3"
}

200response

{
  "audio_data": "SUQzBAAAAAAAI1RTU0UAAAAPAAADTGF2…",
  "audio_format": "mp3",
  "speech_marks": { "chunks": [
    { "value": "სტუმრებს!", "start_time": 1053, "end_time": 2077 },
    …
  ] }
}

One endpoint for every voice. The values that change are highlighted.

Measured

Preferred by listeners, faster under load.

English, against Simba 3.2

63%

win rate against Simba 3.2 in blind side-by-side listening, all content (95% CI 60-67%)¹

1.8×

Simba 3.2’s speech throughput per serving instance²

72 ms

time to first audio for 95% of streaming requests, under production load³

Multilingual, against Simba 3.0 Multilingual

62%

win rate against Simba 3.0 Multilingual in blind side-by-side listening, six locales (95% CI 56-67%)⁴

1.4×

Simba 3.0 Multilingual’s speech throughput per serving instance, on average⁵

languages in one model, against 6 for Simba 3.0 Multilingual

Blind pairwise listening tests: listeners hear the same text in the same voice from two systems, in random order, and choose the one they prefer. Nearly 65,000 judgments from nearly 6,000 US and UK English listeners across 10 content domains. Win rates are derived from Bradley-Terry (Elo-style) ratings fitted to all judgments, excluding ties, with 95% confidence intervals from resampling listeners.
Speech synthesized per second by one serving instance, with both models served in their production configuration (word timing marks on) on the same hardware and the same production text mix.
95th-percentile time from request to first audio while streaming, measured under the same conditions as note 2.
Blind pairwise listening tests in German, Spanish (Spain and Mexico), French, Italian and Portuguese (Brazil), judged by native-speaker listeners: nearly 24,000 judgments from over 2,000 listeners. Same method as note 1; per-locale win rates from a separate fit in each locale.
Speech synthesized per second by one serving instance, with both models served side by side on the same hardware in their production configuration (word timing marks on), with the same prompts; averaged over a standard prompt set.

Results as of October 2026, from the Simba 3.5 fact sheet.

Read the fact sheet