How do i do speech recognition of audio with sensitive content? | ElevenLabs | Page 1

dense geodeBOT Mar 5, 2026, 6:32 AM

#

AI-generated Answer for c.q.q

Short answer
Use Scribe v2 (batch or realtime) and enable Zero Retention Mode plus entity detection.

What to do

Batch STT (files): Call POST /v1/speech-to-text with:
- model_id=scribe_v2
- enable_logging=false (Zero Retention Mode; enterprise-only) to prevent audio/transcript logging for this request
- optional entity_detection to flag sensitive items for redaction (e.g., 'pii', 'phi', 'pci', 'offensive_language' or 'all')
  Example (cURL):
  curl -X POST "https://api.elevenlabs.io/v1/speech-to-text?enable_logging=false"
  -H "xi-api-key: YOUR_API_KEY"
  -H "Content-Type: multipart/form-data"
  -F model_id="scribe_v2"
  -F file=@audio.wav
  -F entity_detection="pii"
  This disables log/transcript storage for the call and returns detected entities with text spans you can redact. (elevenlabs.io)
Realtime STT (WebSocket): Connect to
wss://api.elevenlabs.io/v1/speech-to-text/realtime?model_id=scribe_v2_realtime&enable_logging=false
Authenticate with xi-api-key header (or a single-use token). This also uses Zero Retention Mode per session. (elevenlabs.io)
Zero Retention Mode details: Enterprise feature that prevents storage of STT audio input and text output; enable per request via enable_logging=false. The product matrix lists Speech to Text as eligible. (elevenlabs.io)
Detecting/redacting sensitive data: Use entity_detection with Scribe v2 to locate PII/PHI/PCI and more, with timestamps/character offsets for redaction workflows. (Incurs additional cost.) (elevenlabs.io)
Compliance note (HIPAA/PHI): If you need HIPAA support, you must have a signed BAA before HIPAA-related deployments; ElevenLabs Agents with Zero Retention Mode is the HIPAA-eligible path. (elevenlabs.io)

References

STT overview and HIPAA note: Transcription docs. (elevenlabs.io)
STT batch API (enable_logging, entity_detection): Create transcript. (elevenlabs.io)
STT realtime API (enable_logging): WebSocket reference. (elevenlabs.io)
Zero Retention Mode (enterprise; STT eligible): Docs. (elevenlabs.io)
Entity detection cookbook (PII/PHI examples): Docs. (elevenlabs.io)

If you need help enabling enterprise features like Zero Retention Mode or conversation-history redaction within Agents, contact ElevenLabs Sales per

abstract pasture Mar 5, 2026, 6:34 AM

#

term pf service allow?

dense geodeBOT Mar 5, 2026, 6:34 AM

#

AI-generated Answer for c.q.q

Short answer: By default, no—you must not submit PHI or other regulated sensitive data to ElevenLabs unless you have the proper enterprise agreements and controls enabled.

What the Terms allow (and don’t):

PHI: Not allowed as input unless you have an executed HIPAA Business Associate Agreement (BAA). The Terms explicitly state “No PHI … except as permitted by an executed HIPAA BAA.” (elevenlabs.io)
Other regulated identifiers (financial account numbers, government IDs, etc.): For Conversational AI/Agents, you must not provide this “Prohibited Data” unless ElevenLabs has expressly agreed in writing to meet those requirements. (elevenlabs.io)
Prohibited content/use: Independent of recognition vs. generation, uses that create/promote child sexual content, illegal activity, deceptive/impersonation, election interference, or violent/hateful harassment (outside narrow fictional/news contexts) are disallowed. Review the Prohibited Use Policy before processing such material. (elevenlabs.io)

How to handle sensitive audio if you have the right agreements:

HIPAA workflows: ElevenLabs Agents is HIPAA‑eligible for customers who (1) execute a BAA and (2) enable Zero Retention Mode; in that configuration, handling PHI is permitted. (elevenlabs.io)
Data retention control: Zero Retention Mode (Enterprise) deletes request/response data (including STT audio input and text output) after processing. (elevenlabs.io)
Redaction support: If permitted to process, STT “Entity detection” can flag PII/PHI (e.g., SSNs, credit cards, medical terms) for downstream redaction. (elevenlabs.io)

Data governance notes:

Personal data in your content is governed by the ElevenLabs DPA when applicable; review and incorporate it into your compliance posture. (elevenlabs.io)
General safety/abuse enforcement and reporting are described on the Safety page. (elevenlabs.io)

If you do not have an executed BAA (and written approval where required), do not upload audio containing PHI or other regulated identifiers to ElevenLabs services. (elevenlabs.io)

Sources:
ElevenLabs Terms of Service (non-EEA)
ElevenLabs — Discover our Conversational AI Terms | ElevenLabs
ElevenLabs Prohibited Use Policy
HIPAA | ElevenLabs Documentation
Zero Retention Mode (Enterprise) | ElevenLabs Documentation
Entity detection | ElevenLabs Documentation
ElevenLabs — Discover our Data Processing Addendum (DPA)
Safety
ElevenLabs Terms of Service (non-EEA)

abstract pasture Mar 5, 2026, 6:37 AM

#

why you told me wrong info before
And so what can individual user do

dense geodeBOT Mar 5, 2026, 6:37 AM

#

AI-generated Answer for c.q.q

You’re right—my earlier reply wasn’t verified against ElevenLabs’ official policy. Here’s the correct, source-backed guidance and what an individual user can do.

What’s actually allowed

ElevenLabs’ Prohibited Use Policy applies to both Inputs (your uploaded audio) and Outputs. It bans, among other things, content threatening child safety; illegal activity; deceptive impersonation; election-related campaigning; and violent/hateful/harassing material outside of fictional or news-reporting contexts. Enforcement uses automated and human review. (elevenlabs.io)
ElevenLabs may moderate Inputs/Outputs and, in some cases, share content with third parties to support safety and moderation initiatives. (elevenlabs.io)
Free users may use the Services only for non‑commercial purposes; all users must follow the Prohibited Use Policy. (elevenlabs.io)

Handling sensitive or regulated audio with Speech‑to‑Text (Scribe)

Scribe v2 supports “Entity detection” to automatically flag categories like PII/PHI/PCI (e.g., names, DOB, SSN, credit cards) so you can redact them from transcripts. (help.elevenlabs.io)
Data retention: by default ElevenLabs retains some data to operate and improve services; content may be used for safety/moderation as noted above. Enterprise customers can request Zero Retention Mode, which covers STT audio input and text output; for others, normal retention applies. (elevenlabs.io)

What an individual user can do today

Ensure your use case complies with the Prohibited Use Policy before uploading audio (especially around child safety, illegal activity, impersonation, electioneering, or non‑fiction violent/hateful/harassing content). (elevenlabs.io)
If your audio may include sensitive data, enable Scribe’s entity_detection and remove/redact flagged spans before storing or sharing transcripts. (elevenlabs.io)
Proactively delete transcripts you don’t need via the API: DELETE /v1/speech-to-text/transcripts/:transcription_id. (elevenlabs.io)
If you require zero‑retention processing, note it’s available to enterprise accounts; individuals would need to upgrade or work with an enterprise admin. (elevenlabs.io)
If you’re on the Free plan, use the Services only for non‑commercial purposes. (elevenlabs.io)

If you want, tell me your exact use case and I’ll map the specific API options (e.g., entity types to detect) and policy considerations with citations.

Sources:
ElevenLabs Prohibited Use Policy
ElevenLabs — Read our Privacy Policy
ElevenLabs Terms of Service (non-EEA)
What is Speech to Text? – ElevenLabs
ElevenLabs — Read our Privacy Policy
ElevenLabs Prohibited Use Policy
Entity detection | ElevenLabs Documentation
Delete transcript | ElevenLabs Documentation
Zero Retention Mode (Enterprise) | ElevenLabs Documentation
ElevenLabs Terms of Service (non-EEA)

abstract pasture Mar 5, 2026, 6:44 AM

#

so you are locking user's ability to deal with their own personal sensitive content behind payment gated

dense geodeBOT Mar 5, 2026, 6:44 AM

#

AI-generated Answer for c.q.q

Short answer: No. You can transcribe audio that contains personal/sensitive information on self‑serve plans (including Free). What’s paid are specific add‑ons and enterprise privacy options.

Facts from the docs:

Base Speech‑to‑Text (Scribe v2) is available to self‑serve users; the public pricing page lists Speech‑to‑Text under all plan tiers. (elevenlabs.io)
“Entity detection” (automatic detection of PII/PHI/PCI like names, SSNs, medical conditions, credit cards) is optional, API‑only, and billed as an add‑on. The help article notes it’s API‑only and increases cost; the docs state it carries an additional charge and provide the supported entity list; API pricing shows a separate “Entity detection (per hour)” line. (help.elevenlabs.io)
HIPAA‑eligible use requires a BAA via Sales (i.e., enterprise contracting). (elevenlabs.io)
Zero Retention Mode (immediate deletion of request/response data for especially sensitive workflows) is an Enterprise‑only feature. (elevenlabs.io)
Your content must still comply with the Prohibited Use Policy (e.g., illegal or harmful content is disallowed), but processing your own personal data is not categorically banned. (elevenlabs.io)

What an individual user can do today:

Transcribe sensitive audio normally with Scribe v2 on self‑serve plans. (elevenlabs.io)
If you want automatic flagging/redaction, enable entity_detection in the STT API; budget for the add‑on rate shown on the API pricing page. (elevenlabs.io)
If you require zero‑retention processing and/or a HIPAA BAA, that’s only offered on Enterprise. (elevenlabs.io)

References:

What is Speech to Text? (Scribe v2; entity detection is API‑only and costs more) (help.elevenlabs.io)
STT capability docs (additional cost; HIPAA BAA via Sales) (elevenlabs.io)
Entity detection cookbook (PII/PHI/PCI details; add‑on) (elevenlabs.io)
API pricing (separate line item for entity detection) (elevenlabs.io)
Zero Retention Mode (Enterprise‑only) (elevenlabs.io)
Prohibited Use Policy (elevenlabs.io)
Plan overview with STT listed under Free and paid

abstract pasture Mar 5, 2026, 6:46 AM

#

it won't violate tos?

dense geodeBOT Mar 5, 2026, 6:46 AM

#

AI-generated Answer for c.q.q

Short answer: It depends on the content and context.

It should be OK (i.e., not a ToS violation) to transcribe “sensitive” audio if all of the following are true:

The content does not fall under the Prohibited Use Policy (e.g., no sexual content involving minors; no illegal activity; no real‑world violent, hateful, or harassing material outside of fiction/news reporting; no deceptive impersonation). (elevenlabs.io)
You have the necessary rights/consents to the audio you upload and you do not violate others’ privacy or IP rights. (elevenlabs.io)
You do not include HIPAA‑protected health information unless you have a signed BAA with ElevenLabs (required for HIPAA use). (elevenlabs.io)

It will violate ToS if, for example:

The audio includes prohibited content/categories noted above, or is used to facilitate illegal/deceptive conduct (including unauthorized impersonation). (elevenlabs.io)
You lack rights to the material or process others’ private data without appropriate legal basis/consent. (elevenlabs.io)
You process PHI without a BAA. (elevenlabs.io)

Additional note: ElevenLabs reserves the right to moderate inputs/outputs and take enforcement action even if you believe the content is compliant. (elevenlabs.io)

If you can share the exact type of “sensitive content,” I can map it to the specific policy clauses.

Sources:
ElevenLabs Prohibited Use Policy
ElevenLabs Terms of Service (non-EEA)
ElevenLabs Terms of Service (non-EEA)
ElevenLabs Prohibited Use Policy
ElevenLabs Terms of Service (non-EEA)
ElevenLabs Terms of Service (non-EEA)
ElevenLabs — Read our Privacy Policy

abstract pasture Mar 5, 2026, 6:53 AM

#

is "starting a revolution in real world" against prohibited use policy?

dense geodeBOT Mar 5, 2026, 6:53 AM

#

AI-generated Answer for c.q.q

Short answer: Yes—if it’s calling for real‑world violent action.

Details:

ElevenLabs prohibits content that “threatens, incites, or promotes violence,” including extremism or terrorism. Calls to start a real‑world revolution that entail or encourage violence fall under this ban. Fictional or news‑reporting contexts are exempted. See “Do not create violent, hateful, or harassing material outside of fictional contexts,” section 8(a) and the noted exceptions. https://elevenlabs.io/use-policy
ElevenLabs also states it may refer material posing real‑world risk of harm to law enforcement. https://elevenlabs.io/safety

Non‑violent, general political discussion is not categorically banned, but election‑context political campaigning is restricted (section 6). https://elevenlabs.io/use-policy

abstract pasture Mar 5, 2026, 6:55 AM

#

so how do i transcribe audio of such content?

dense geodeBOT Mar 5, 2026, 6:55 AM

#

AI-generated Answer for c.q.q

Short answer: Use Scribe v2 via the Speech to Text API. You may transcribe “sensitive” topics provided your use does not violate ElevenLabs’ Prohibited Use Policy (e.g., content that threatens, incites, or promotes violence, extremism, or terrorism is prohibited outside of clear fictional/news reporting contexts). (elevenlabs.io)

How to transcribe it safely and correctly

Choose batch or realtime:
- Batch (files): POST /v1/speech-to-text with model_id=scribe_v2. Supports audio/video, up to 3 GB and 10 hours; multichannel (up to 5 channels) and webhooks for async processing. (elevenlabs.io)
- Realtime (streaming): WebSocket wss://api.elevenlabs.io/v1/speech-to-text/realtime with model_id=scribe_v2_realtime. (elevenlabs.io)
Detect and handle sensitive info in transcripts:
- Enable entity detection to mark PII/PHI/PCI or offensive_language so you can redact it in your pipeline:
  - entity_detection=["pii","offensive_language"] (or "all"). (elevenlabs.io)
- For Agent conversation storage, enterprise customers can enable automatic redaction of sensitive entities in stored transcripts/audio. (elevenlabs.io)
Data handling (if your audio contains personal/sensitive data):
- Enterprise zero‑retention: set enable_logging=false on the STT request to use Zero Retention Mode (logs/transcript storage disabled for that call). Also see the Zero Retention Mode overview. Otherwise, standard retention applies per the Privacy Policy. (elevenlabs.io)

Minimal batch example (Python SDK)

The quickstart shows a working call; add entity detection if you need redaction downstream:
- elevenlabs.speech_to_text.convert(file=..., model_id="scribe_v2", language_code="eng", diarize=True, tag_audio_events=True) (elevenlabs.io)
- elevenlabs.speech_to_text.convert(..., entity_detection=["pii","offensive_language"]) (elevenlabs.io)

Key capabilities and limits

90+ languages, word‑level timestamps, speaker diarization (up to 32 speakers), dynamic audio tagging. (help.elevenlabs.io)
Supported formats include common audio/video types (e.g., MP3, WAV, MP4). Size ≤3 GB, duration ≤10 hours; webhooks available for async. (elevenlabs.io)

Policy reminder

Ensure your content and usage comply with the Prohibited Use Policy (e.g., no material that threatens, incites, or promotes violence; no terrorism/extremism content outside fictional/news contexts). (elevenlabs.io)

If you want, I can provide a concrete cURL or code sample tailored to your file (batch or realtime) using the flags above. (elevenlabs.io)

abstract pasture Mar 5, 2026, 6:57 AM

#

didn't we already talked about content inside prohibited use?

dense geodeBOT Mar 5, 2026, 6:57 AM

#

AI-generated Answer for c.q.q

Yes—we did. For reference, ElevenLabs’ Prohibited Use Policy (last updated September 3, 2025) governs both Inputs and Outputs and applies on and off the platform. It bans, among other things: child sexual content; illegal activities; harmful impersonation/voice cloning without consent; election-related deception and voter suppression; harassment, discrimination, self-harm promotion, and terrorism/violence (with limited exceptions for purely fictional contexts or news reporting). Enforcement can include content removal, suspension/ban, and referrals to law enforcement. (elevenlabs.io)

If you want me to map a specific transcription scenario to the policy, share the details and I’ll point to the exact clause. (elevenlabs.io)

Sources:
ElevenLabs Prohibited Use Policy
ElevenLabs Prohibited Use Policy

abstract pasture Mar 5, 2026, 7:27 AM

#

so how do i transcribe those content?

dense geodeBOT Mar 5, 2026, 7:27 AM

#

AI-generated Answer for c.q.q

Short answer: Use Scribe v2 via the Speech-to-Text API (file upload/URL) or Scribe v2 Realtime (WebSocket). Ensure your audio complies with the Prohibited Use Policy; inputs are covered by that policy.

What’s allowed vs not

ElevenLabs’ Prohibited Use Policy applies to both Inputs and Outputs. It prohibits material that threatens child safety; illegal activity; and violent, hateful, or harassing material—except when purely fictional or part of reporting on newsworthy activity by third parties. If your audio includes calls to real‑world violence (e.g., “start a revolution”) outside those carve‑outs, do not upload it. (elevenlabs.io)

How to transcribe (batch/file)

Endpoint: POST /v1/speech-to-text with model_id=scribe_v2, providing either file (multipart) or cloud_storage_url. Optional: language_code, keyterms (up to 100), entity_detection (e.g., 'pii', 'phi', 'pci', 'offensive_language'), and webhook for async delivery. Example cURL is on the page. (elevenlabs.io)

How to transcribe (realtime/streaming)

WebSocket: wss://api.elevenlabs.io/v1/speech-to-text/realtime with model_id=scribe_v2_realtime. Send input_audio_chunk frames; you’ll receive partial_transcript and committed_transcript events (optionally with word‑level timestamps). Authenticate with xi-api-key or a single‑use token. (elevenlabs.io)
To authenticate client‑side safely, create a single‑use token: POST /v1/single-use-token/realtime_scribe, then pass it as the token query parameter when opening the WebSocket. (elevenlabs.io)

Privacy and sensitive data handling

You can set enable_logging=false on both endpoints to use zero‑retention mode; note this mode is available to Enterprise customers only. (elevenlabs.io)
The Speech‑to‑Text capability docs also cover formats, webhooks, and operational limits (e.g., large files and multichannel); refer there if you need specifics. (elevenlabs.io)

If you’re unsure whether your specific content is permitted under the Prohibited Use Policy, don’t upload it until you’ve reviewed that policy in detail. (elevenlabs.io)

Sources:
ElevenLabs Prohibited Use Policy
Create transcript | ElevenLabs Documentation
Realtime | ElevenLabs Documentation
Create Single Use Token | ElevenLabs Documentation
Create transcript | ElevenLabs Documentation
Transcription | ElevenLabs Documentation
ElevenLabs Prohibited Use Policy

abstract pasture Mar 5, 2026, 7:49 AM

#

should i find alternative service for content that cannot be processed within elevenlabs? Any recc?

dense geodeBOT Mar 5, 2026, 7:49 AM

#

AI-generated Answer for c.q.q

Short answer: If your content falls under ElevenLabs’ Prohibited Use Policy, you must not process it on ElevenLabs (this applies to Text-to-Speech, Speech-to-Text, ElevenReader, etc.). ElevenLabs’ docs do not provide or endorse recommendations for third‑party alternatives.

Key points to check before switching:

Prohibited content: See the Prohibited Use Policy for what cannot be processed and note the carve‑outs for purely fictional contexts or reporting on newsworthy activity. (elevenlabs.io)
Speech-to-Text is subject to the same policy: The Speech to Text Terms expressly bind STT to the Prohibited Use Policy. (elevenlabs.io)
Political-figure voice cloning: “No‑Go Voices” (e.g., election‑related political figures) are blocked by design. (help.elevenlabs.io)
HIPAA/PHI: If your “sensitive content” is protected health information, ElevenLabs requires a signed BAA before using STT for HIPAA‑regulated workflows. (elevenlabs.io)

If you’re unsure whether your specific use is allowed or if something was blocked in error, contact ElevenLabs Support for a case review. (help.elevenlabs.io)

Sources:
ElevenLabs Prohibited Use Policy
Speech to Text Terms
What are No-Go Voices? – ElevenLabs
Transcription | ElevenLabs Documentation
How can I contact support? – ElevenLabs

abstract pasture Mar 5, 2026, 7:50 AM

#

any outside service you can recommend?

dense geodeBOT Mar 5, 2026, 7:50 AM

#

AI-generated Answer for c.q.q

ElevenLabs’ official documentation does not provide or endorse recommendations for third‑party transcription or moderation services. If your use case is permitted under our policies, you can transcribe it with ElevenLabs Speech‑to‑Text; if it falls under prohibited content/uses, it must not be processed on ElevenLabs, and we can’t advise workarounds or alternatives. (elevenlabs.io)

Check what’s allowed: Prohibited Use Policy and Terms of Service. (elevenlabs.io)
If allowed, implement via Speech‑to‑Text docs/quickstart. (elevenlabs.io)
For edge cases or clarifications, contact Support. (help.elevenlabs.io)

Sources:
ElevenLabs Prohibited Use Policy
ElevenLabs Prohibited Use Policy
Transcription | ElevenLabs Documentation
How can I contact support? – ElevenLabs

abstract pasture Mar 5, 2026, 7:51 AM

#

the issue is it isn't allowed

dense geodeBOT Mar 5, 2026, 7:51 AM

#

AI-generated Answer for c.q.q

Short answer: Using ElevenLabs to process audio is disallowed if the audio falls under our Prohibited Use Policy. The Policy applies to both inputs you upload (e.g., source audio) and outputs you generate (e.g., transcripts).

Key points you can rely on:

The Prohibited Use Policy applies to “any Inputs you provide and Outputs you create.” This includes uploads for transcription. (elevenlabs.io)
Prohibited categories include (non‑exhaustive): threats to child safety; illegal activity; and creating, promoting, or facilitating violent, extremist, hateful, or harassing material. There is an explicit carve‑out that the “violent/hate/harassment” section does not apply to purely fictional contexts or reporting on newsworthy activity by third parties. (elevenlabs.io)
If your use involves health information and you require HIPAA compliance, you must sign a BAA with ElevenLabs before proceeding. (elevenlabs.io)
For handling sensitive data, you can enable Zero Retention Mode so content processed by our models isn’t retained on ElevenLabs servers. (elevenlabs.io)
Our Safety page also links to and summarizes these enforcement standards. (elevenlabs.io)

If your audio includes prohibited content (for example, real‑world incitement to violence or illegal material), you cannot process it with ElevenLabs—even for transcription. If it is lawful and not within prohibited categories (or it falls under the noted exceptions such as fictional or news reporting contexts), you may use Speech‑to‑Text subject to all policy and legal requirements. (elevenlabs.io)

Sources:
ElevenLabs Prohibited Use Policy
ElevenLabs Prohibited Use Policy
Transcription | ElevenLabs Documentation
Data residency | ElevenLabs Documentation
Safety
ElevenLabs Prohibited Use Policy

abstract pasture Mar 5, 2026, 7:56 AM

#

let say if it is audio recording of advocating the rebellion against current PRC government, which is illegal in PRC

dense geodeBOT Mar 5, 2026, 7:56 AM

#

AI-generated Answer for c.q.q

Short answer: If the recording advocates or incites violent rebellion, you should not process it on ElevenLabs. Our Prohibited Use Policy bans using the Services (including any Inputs you provide and Outputs you create) to create, distribute, or engage in material that threatens, incites, or promotes violence, extremism, or terrorism. An explicit exception exists only for purely fictional contexts or when the use is part of reporting on newsworthy activity by third parties.

Separately, illegal behavior is prohibited, and ElevenLabs may remove content and suspend accounts for policy or law violations; we may also refer criminal or other illegal activity to law enforcement. You are responsible for complying with all laws applicable to your use.

Therefore:

Direct advocacy of violent rebellion: disallowed.
Journalistic reporting on third‑party speech (newsworthy context): may be permitted under the narrow exception, but enforcement is at ElevenLabs’ discretion.

If you are unsure whether your specific use fits the “news reporting” exception, do not upload it and seek clarification via the Help Center.

abstract pasture Mar 5, 2026, 7:58 AM

#

Not elevenlabs, then where

dense geodeBOT Mar 5, 2026, 7:58 AM

#

AI-generated Answer for c.q.q

ElevenLabs can’t be used for content that promotes or incites violence, extremism, illegal activity, or election-related campaigning. “Advocating rebellion” would fall under prohibited uses (e.g., violent threats/extremism/terrorism; illegal behavior; political campaigning), except in purely fictional or news-reporting contexts. Violations can lead to bans and, in serious cases, referral to law enforcement. (elevenlabs.io)

ElevenLabs documentation does not endorse or recommend third‑party services; references to third‑party products do not constitute endorsement. Accordingly, we can’t advise “where else” to process prohibited content. (elevenlabs.io)

If your use case is lawful and compliant with the policy, you can transcribe with ElevenLabs Scribe (Speech to Text) in 90+ languages via web or API. (help.elevenlabs.io)

Please review the Prohibited Use Policy and Safety page to confirm whether your specific material is allowed before proceeding. (elevenlabs.io)

Sources:
ElevenLabs Prohibited Use Policy
ElevenLabs Terms of Service (non-EEA)
What is Speech to Text? – ElevenLabs
ElevenLabs Prohibited Use Policy

abstract pasture Mar 5, 2026, 8:00 AM

#

what if the language of the audio track is Taiwanese Min

dense geodeBOT Mar 5, 2026, 8:00 AM

#

AI-generated Answer for c.q.q

ElevenLabs Speech-to-Text (Scribe v2) does not list Taiwanese Min (Min Nan/Hokkien; ISO 639-3 “nan”) among its supported languages. The documented Sinitic variants currently include Mandarin Chinese (zho) and Cantonese (yue), but not Min Nan. Please refer to the official supported-languages list for the latest status. (elevenlabs.io)

General language-support guidance is also in the Help Center. (help.elevenlabs.io)

Sources:
Transcription | ElevenLabs Documentation
Which languages does Speech to Text support? – ElevenLabs

abstract pasture Mar 5, 2026, 8:02 AM

#

Not elevenlabs, then where

dense geodeBOT Mar 5, 2026, 8:02 AM

#

AI-generated Answer for c.q.q

ElevenLabs does not recommend or support third‑party services. The help center states we cannot assist with other services and offers no external provider list. (help.elevenlabs.io)
If the content is prohibited under our Prohibited Use Policy (e.g., violent threats, extremism/terrorism outside of fictional or news‑reporting contexts), you cannot process it with ElevenLabs. The policy applies to both Inputs and Outputs. (elevenlabs.io)
Speech‑to‑Text (Scribe) is also explicitly subject to the Prohibited Use Policy; using STT for prohibited content is not allowed. (elevenlabs.io)
For clarification on whether a specific use might be permitted (e.g., reporting on newsworthy activity), please contact ElevenLabs Support. (help.elevenlabs.io)

Terms reference: your use must comply with our Terms of Service and the Prohibited Use Policy. (elevenlabs.io)

Sources:
I'm having issues with a third-party service, can you help me? – ElevenLabs
ElevenLabs Prohibited Use Policy
Speech to Text Terms
How can I contact support? – ElevenLabs
ElevenLabs Terms of Service (non-EEA)

abstract pasture Mar 5, 2026, 8:06 AM

#

but you said you don't do Taiwanese

dense geodeBOT Mar 5, 2026, 8:06 AM

#

AI-generated Answer for c.q.q

Clarifying language support:

ElevenLabs does not list Taiwanese Hokkien (Taiwanese Min/Hokkien) as a supported language for Text to Speech. The TTS language list includes Chinese (Mandarin) but not Taiwanese Hokkien. (help.elevenlabs.io)
For Speech to Text (Scribe v2), supported languages include Mandarin Chinese and Cantonese; Taiwanese Hokkien is not listed. (elevenlabs.io)
If you meant Mandarin as spoken in Taiwan, you can use Chinese (Mandarin) in TTS/STT; for the most natural TTS results, pick a voice whose accent/region matches your target language. (elevenlabs.io)

If you need Taiwanese Hokkien specifically, it isn’t officially supported today. (help.elevenlabs.io)

Sources:
What languages do you support? – ElevenLabs
Transcription | ElevenLabs Documentation
Text to Speech | ElevenLabs Documentation
What languages do you support? – ElevenLabs

abstract pasture Mar 6, 2026, 8:27 PM

#

再講多次

dense geodeBOT Mar 6, 2026, 8:27 PM

#

AI-generated Answer for c.q.q

你想我重複邊一部分內容？請告知具體的 ElevenLabs 問題或想重溫的主題（例如文字轉語音、Voice Library、API 用法、計費等），我會按官方文件解答並附上來源。

abstract pasture Mar 6, 2026, 8:28 PM

#

唔該講俾我聽有咩工具可以做到台語文字轉語音

dense geodeBOT Mar 6, 2026, 8:28 PM

#

AI-generated Answer for c.q.q

簡短結論：目前 ElevenLabs 未有正式支援「台語／臺灣閩南語（Hokkien）」的文字轉語音；官方語言清單只列出「中文（CMN，普通話）」，因此無法保證台語輸出。 (help.elevenlabs.io)

補充：

系統會根據你輸入的文字自動偵測語言；而實際的口音與發音風格主要由所選用的聲音決定。若你想要「台灣國語」風格的中文，可選擇適合的中文聲音，但這仍屬中文（普通話），不是台語。 (help.elevenlabs.io)
如需追蹤是否新增語言，請以官方「支援語言」頁為準。 (help.elevenlabs.io)

想確認你需要的是台語（閩南語）還是台灣國語的中文？

Sources:
What languages do you support? – ElevenLabs
What languages can the AI speak? – ElevenLabs
What languages do you support? – ElevenLabs

abstract pasture Mar 6, 2026, 8:30 PM

#

而家十一實驗室嘅工具做唔到嘅話咁點先做到

#How do i do speech recognition of audio with sensitive content?