ElevenLabs v3
In the ElevenLabs voice architecture, v3 is the only component that implements Persian in its language-coverage surface, registered as fas within a 74-language list. For any workload with Persian input, manually selecting this model is an operational requirement, because this vendor faster and cheaper components carry no such coverage.
- 74-language coverage including Persian
- 5,000 character processing ceiling
- Voice cloning with a commercial license
Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. Change log
Why manual selection is required
The ElevenLabs text-to-speech stack is built from three components, and only this one registers Persian in its language-coverage surface. Flash v2.5 covers a 32-language list and Multilingual v2 covers a 29-language list, and Persian is present in neither specification.
The component selected by default is Flash v2.5, because its latency is 75ms, its character processing ceiling is eight times larger, and its cost is lower. For a workload with Persian input this is the wrong choice; the resulting drop in output quality is not a language limitation, it is a component-selection error.
The operating cost of this component
The processing ceiling per call is 5,000 characters, against 40,000 for Flash v2.5. So the one component with Persian coverage is the one that requires manual segmentation of input and reassembly of output. For a video of a few minutes that means several sequential calls followed by an edit pass.
We record this not as a defect but as part of the operating specification to account for before the project pipeline is designed.
Ownership status of the audio output
The voice-cloning subsystem in this vendor architecture is tied to the subscription tier, not to the selected model. The free tier carries neither cloning capability nor a commercial license. Instant cloning becomes available at the Starter tier for $6, and higher-fidelity cloning at the Creator tier for $22.
As a result, output generated on the free tier is unlicensed for commercial use, regardless of how acceptable its quality is judged. This constraint is stated in the vendor own pricing specification and is not flagged anywhere else.
A gap in this model voice architecture
No dedicated Persian voice is defined in this architecture. ElevenLabs registers Persian in its language-coverage surface, but no voice trained specifically for Persian is named in its specification. In practice, Persian input is processed through a voice built on another language, and the resulting accent quality is a property that has to be evaluated by direct listening; that evaluation has not yet been performed on this component.
Using it from Iran
This column carries a date and says how it was checked, because most listicles guess it. Where we have only read a vendor policy page, the note below says exactly that.
- Reachable
- blocked
- Payment
- no working route
- Free tier
- yes
- Measured on
How we checked: ElevenLabs terms-of-use document names Iran explicitly in its export-control clause: the user must warrant they are not located in a country under US economic sanctions, and Iran is listed there. We extracted this constraint from the text of that clause itself, not from a direct network measurement.
Documented strengths
- Persian is registered as fas within this component 74-language coverage surface, a status unique across this vendor entire voice stack
- The voice-cloning subsystem operates from a single input sample, with an active commercial license from the Creator tier upward
- The highest supported-language count among all three text-to-speech components from this vendor
Known weaknesses and limits
- Its processing ceiling is 5,000 characters per call, an eighth of Flash v2.5 capacity; long text has to be segmented in the input pipeline.
- No dedicated voice for Persian is defined in this component architecture, so Persian processing runs through a voice built for a different language.
- The Persian accent quality of this component has not been measured; this document only confirms the language presence at the specification level.
- The ElevenLabs terms-of-use document names Iran explicitly in its export-control clause, and no official access route from Iran is defined in this service access architecture.
Technical verdict
For any workload with Persian input in the ElevenLabs stack, this component is the only option to select in the pipeline configuration. The pipeline design has to account for its small processing ceiling and for the free tier granting no commercial usage rights.
Frequently raised questions, with documented answers
Why was my Persian output quality low
The most likely cause is that the configuration was set to the Flash v2.5 component. That component covers a 32-language list and Persian is not registered in it. For Persian input, v3 has to be selected manually in the configuration.
What does a 5,000 character processing ceiling per call mean
It means that is the maximum text that can be sent in a single API call. For longer text you make several sequential calls and concatenate the resulting audio files in a later step.