ElevenLabs Voice Cloning and Voice Design API vs Gradium Voice Cloning and Voice Design

ElevenLabs Voice Cloning and Voice Design API 84

Graded for voice cloning, with consent and misuse controls in place of the read-only line and training and retention of voice data in place of the prompt-injection line, since the API returns audio and IDs rather than third-party text. API keys can be limited to chosen endpoints, given a credit quota, user keys can expire after 15 minutes to 30 days, service-account keys exist and accept an expiry of up to 90 days since 26 October, keys can be limited to IP ranges, and leaked keys are disabled through GitHub secret scanning (30). PVC is own-voice only and needs a spoken verification matched to the samples, IVC relies on an attestation, high-risk and celebrity voices are blocked, and an AI speech classifier and C2PA support are public (14 of 20). Training on content is on by default with a self-serve opt-out, and the privacy policy caps voice data retention at 3 years after the last interaction (10 of 15). The History API lists each generation with voice, model, characters and date (15). security.txt valid to 2027-10-06 (re-read on 1 November, contact elevenlabs.io/security/report) with a disclosure programme, SOC 2 Type II, and a trust centre at compliance.elevenlabs.io, no paid bug bounty found (15 of 20).

Gradium Voice Cloning and Voice Design 48

Graded for voice cloning as the other listings in this category, with consent and misuse controls in place of the read-only line and training and retention of voice data in place of the prompt-injection line. API keys per organisation in a header, revocable and able to expire, with no scopes found (20 of 30). No consent or verification field and no watermark. The Terms make the client obtain the consent of anyone whose voice is a sample and bar impersonation without explicit authorisation, and the API FAQ says explicit consent is required (6 of 20). The Terms grant Gradium a licence to use Client Data to improve its services and train its models, with no opt-out found, and zero data retention (paid plans) excludes voice samples, so samples are kept and may be used for training (2 of 15). Organisation roles and a member list, with no per-call audit log found (5 of 15). SOC 2 Type II and ISO 27001:2022 audited by A-LIGN, a trust centre and security@gradium.ai. No security.txt (404) and no bug bounty found (15 of 20). 20+6+2+5+15 = 48.

ElevenLabs Voice Cloning and Voice Design API 88

Creating a clone returns only voice_id and requires_verification, and GET /v2/voices can be narrowed to personal or cloned voices (25). Cursor pagination with page_size up to 100 and filters by type, category, fine-tuning state and labels (20). Errors carry a type, a named code and a param for validation failures, with recovery hints on many codes such as voice_not_found, invalid_cursor and concurrent_modification, though the two concurrency codes in different pages disagree (18 of 20, up from 15). No idempotency key, PVC exposes fine_tuning_state to poll (10 of 20). Two required fields for an instant clone, and official Python and TypeScript SDKs among others (15).

Gradium Voice Cloning and Voice Design 67

Voice objects are compact metadata (20 of 25). GET /voices/ takes skip and limit (default 100) and since 23 October filters by locale, gender and use_case (20). Validation errors are structured 422 bodies, while synthesis errors arrive as plain-text HTTP 500 with a numeric code to parse (12 of 20). No idempotency key. Saving a design candidate twice answers 409, and the clone call has none (5 of 20). Three required fields with sensible defaults, and one official SDK, in Python (10 of 15). 20+20+12+5+10 = 67.

ElevenLabs Voice Cloning and Voice Design API 83

Closed service with clear terms, SDKs under MIT (15 of 30). Privacy policy updated 9 October 2026 (unchanged on 1 November), still with a 3 year retention cap for voice data and a training opt-out under Data use, a DPA, subprocessors through the compliance portal and a training opt-out that matches the Terms, but the Terms (last updated 9 October 2026, unchanged on 1 November) also take a perpetual, irrevocable licence to User Voice Models while promising deletion on request, which sit uneasily together (25 of 30). Dated deprecations in the changelog, no written notice policy found (15 of 20). Data locations named (United States, the Netherlands, Singapore) and Enterprise residency in the EU, India and Singapore (20).

Gradium Voice Cloning and Voice Design 63

Closed service with clear terms, version of 19 October 2026, naming Gradium SAS with its Paris registration (15 of 30). The Terms allow training on Client Data, zero data retention excludes voice samples, the privacy policy gives retention criteria without periods, and a DPA is published as an exhibit. The statements agree but give no retention period for samples (15 of 30). Deprecations appear in dated release notes, with no written notice policy (10 of 20). EU and US residency is documented and selectable from the XS plan. Sub-processors sit in the trust centre, not read (12 of 20). 15+15+10+12 = 52.

ElevenLabs Voice Cloning and Voice Design API 85

Python and JavaScript SDK 2.75.0, tagged on 27 October 2026 (23:02 UTC), after 2.74.0 and 2.74.1 on 20 October and 2.73.1 on 17 October (30). Changelog entries weekly, on 5, 12, 19 and 26 October, and seven Python SDK releases in October (20). Public changelog and a help centre, and SDK issues get labels, but the ten newest open issues showed no maintainer replies on 1 October, per the 1 October check, not re-read since (10 of 15). Current official SDKs (15). The Python SDK supports 3.8 and later and shipped five days before this check (10).

Gradium Voice Cloning and Voice Design 70

Read as a model, on the lines the other hosted cloning listings use (recency 30, dated releases 10, deprecation notice 10, support 15, SDKs 15, health 10). Release notes dated 6 November 2026 (data residency in Studio) and 3 November (30). Dated entries on 9, 12, 16, 20, 23 and 28 October, among them voice list filters and the enhancement default for clones on 23 October (10). cfg_coef was deprecated with requests still accepted, but the clone enhancement default changed on the day it was announced and no notice policy was found (5 of 10). Public release notes, support@gradium.ai and a Discord server, with no evidence read of how they answer (10 of 15). gradium on PyPI at 0.6.4 (4 September 2026), with no SDK in a second language (10 of 15). The SDK repository was not cloned (5 of 10). 30+10+5+10+10+5 = 70.

ElevenLabs Voice Cloning and Voice Design API 73

Read as a hosted service, as on 1, 6 and 17 October. Status page at status.elevenlabs.io (incident.io) with component history (20). Re-read on 1 November from the history page, which dates each component's impact. The incidents feed was re-posted on 21 October and leaves these incidents out, so the record rests on the history page. The incident the earlier notes counted as a clone-path major, elevated TTS and STT error rates on 3 August, ran 06:39 to 07:13 UTC (34 minutes) at degraded performance, not about 2 hours, so it is minor, as the elevenlabs-tts listing read it. In the 90 days from 3 August the clone path also saw degraded TTS and STT latency from 15:15 to 17:53 UTC on 26 August (2 hours 38 minutes, not the 8 hours the 17 October note gave), about 3 hours of TTS latency on 6 October, 11 minutes of TTS failures on 12 October, Instant Voice Clone creation failures from 03:25 to 04:44 UTC on 13 October (Other APIs, degraded, limited to requests with background noise removal on) and the creative app degraded from 09:19 to 11:01 UTC on 30 October. The partial outages hit Agents, Telephony, Dubbing and STT, including about 5 hours of US timeouts on 13 October on Agents components only. Minor incidents only (20 of 30, was 10). A stricter reading of 13 October as a clone-path major would give 10 here. Concurrency per plan is published, 2 on Free up to 15 on Scale and Business (15). 429s are documented with exponential backoff guidance and a wait-for-completion rule for concurrency, though the help centre names the concurrency code too_many_concurrent_requests and the errors reference concurrent_limit_exceeded (8). No idempotency key or safe-retry guidance for creating a voice (0). Enterprise mentions custom DPA and SLA terms, but no SLA is published (0). IVC, PVC and Voice Design are generally available (10). 20 + 20 + 15 + 8 + 0 + 10 = 73.

Gradium Voice Cloning and Voice Design 63

Graded on the hosted voice API with the hosted lines. Atlassian Statuspage at gradium.statuspage.io with five components, one of them Voice Cloning (20). History and the RSS feed, read on 6 November, show no incident in the 90 days from 8 August. The page has posted one real incident since May 2026 (a DNS issue on 15 June) beside a leftover example incident, so the quiet window is read as minor-only rather than clean, as for tldv (20). Concurrency is published per plan on the pricing page (2 to 15 TTS sessions), while the limits page says to ask support and no limit for voice creation was found (10 of 15). No 429, Retry-After or backoff guidance was found; the errors page says to open a new connection after an error, and voice creation takes a timeout_s (3 of 15). No SLA in the Terms or on the pricing page (0). Instant cloning carries no beta label (10). 20+20+10+3+0+10 = 63.

ElevenLabs Voice Cloning and Voice Design API 91

Public OpenAPI at api.elevenlabs.io/openapi.json, 313 paths per the 6 October check, not re-read on 17 October or 1 November because api.elevenlabs.io's robots.txt now disallows every path (25). llms.txt and Markdown docs, with a .md twin of every page (10). The cloning guides say when to use IVC or PVC and warn that more than 3 minutes of IVC audio can make a clone worse (16 of 20). Inputs are typed, but labels is a free-form map (11 of 15). Plenty of examples, and an errors reference in the API docs lists 10 error types with their HTTP statuses and about 70 named codes, though endpoint pages in the OpenAPI still list only 200 and 422 (14 of 15, up from 10, which read the codes as living only in help-centre articles). Versioned paths (/v1, /v2) and a changelog published several times a month, the latest entries dated 12, 19 and 26 October (15).

Gradium Voice Cloning and Voice Design 83

Public OpenAPI 3.1 file at docs.gradium.ai/api-reference/openapi.json (25). llms.txt and Markdown pages (10). The guides say when to clone, design, enhance or localise a voice, and the custom-voice page explains the enhancement default and how to opt out (14 of 20). The clone body has audio_file, name and language required, a five-value language enum and typed defaults for start_s, enhance and timeout_s (14 of 15). curl and Python examples. The create call documents only 201 and 422, and the errors page covers synthesis, with plain-text 500 bodies (9 of 15). Routes carry no version, the OpenAPI file is 0.1.0, and release notes are dated per entry, newest 6 November 2026 (11 of 15). The API reference FAQ gives a 300-second session where the limits page says 3,000. 25+10+14+14+9+11 = 83.