Inworld Instant Voice Cloning vs VoiceStudio

Inworld Instant Voice Cloning 75

Graded on the REST API, the surface an agent would call. Status page at status.inworld.ai, drawn from GitHub issues, with five components and none for cloning (20). Read through the GitHub API on 6 November, the 90 days from 8 August hold two incidents, 29 minutes of elevated TTS latency on 25 August and 26 minutes of elevated TTS 2.1 errors from a network fault at one GPU site on 28 October, both minor (20). Cloning limits are published per plan, 2 to 20 requests a minute (15). A 429 is documented with exponential backoff and jitter code, and since the 9 October check the rate-limits page says limits over a minute can carry Retry-After. No idempotency key or safe-retry guidance for clone creation (10 of 15, was 9). The pricing page, now read, names an SLA on Enterprise only with no terms or figure, as on inworld-tts (0). Instant cloning carries no Preview or Beta mark (10). 20+20+15+10+0+10 = 75.

VoiceStudio 45

Read with the local-package rubric, as a desktop app with a local MCP server over Streamable HTTP or a stdio shim. Installers for Windows, macOS, Linux, Android and iPhone on each GitHub release, with platforms and runtime setup documented (20). A public GitHub Actions workflow runs pytest on the backend and tests, the frontend tests and Electron tests on every pull request and push to main. Its results couldn't be read, since GitHub's pages are closed to us (15). Open crash and regression issues couldn't be read for the same reason (0, unchecked). Dated 0.x releases in CHANGELOG.md with Changed, Added, Removed and Fixed sections and upgrade notes, but no semver promise at 0.5 (10, the earlier partial, kept). Version 0.5.7, not 1.0 and not declared stable (0). 20+15+0+10+0 = 45.

Inworld Instant Voice Cloning 68

Read as a model, on the lines inworld-tts and cartesia-voice-cloning use (recency 30, dated releases 10, deprecation notice 10, support 15, SDKs 15, health 10). The 9 October draft used the tool lines and gave 75. @inworld/cli 1.5.0, whose MCP server carries clone_voice, on 29 October 2026, after 1.4.0 (13 October), 1.4.1 (23 October), 1.4.2 and 1.4.3 (25 October) (30). Dated TTS release notes on 9 August and 11 September 2026, and none for the 28 October model update or for the Voice API (10). inworld-tts-1 was discontinued on 15 June 2026 with requests routed to newer models, and clones made before 15 August keep a 15-second prompt, with no notice period stated (5 of 10). A public changelog, a Discord server and support@inworld.ai, with no evidence read of how they answer (8 of 15). @inworld/tts 1.1.1 dates from 6 May 2026 (10 of 15). The CLI is current, but the repositories named in its npm metadata aren't public (5 of 10). 30+10+5+8+10+5 = 68.

VoiceStudio 50

Version 0.5.7 released on 26 October 2026, 10 days before this check (30). Four stable releases since 7 August, 0.5.4 (20 September), 0.5.5 (22 September), 0.5.6 (23 September) and 0.5.7 (26 October). The three 0.5.7 betas don't count (20). Issue and pull request replies couldn't be read, since GitHub's pages are closed to us (0, unchecked). No official MCP registry entry was found on a readable page (0). Package health not established. CI is public and 0.5.7 updated bundled dependencies past published advisories, but the CI results couldn't be read (0). 30+20+0+0+0 = 50.

Inworld Instant Voice Cloning 63

Closed service with clear terms naming Theai, Inc. dba Inworld AI, read on 6 November (unread on 9 October because of the Content-Signal line), and MIT client packages on npm (15 of 30, was 8). The Zero Data Retention page is specific that cloning samples are retained to build the voice, but gives no period, and the Terms' no-training clause and the Privacy Notice's training use sit side by side without a reconciling sentence, as on inworld-tts (14 of 30). Dated notices in the docs, inworld-tts-1 discontinued on 15 June 2026 and the 30-second clone limit from 15 August, with release stages defined and no written notice period (12 of 20). US by default with isolated EU and India regions on Enterprise. No sub-processor list was found in the docs (10 of 20). 15+14+12+10 = 51.

VoiceStudio 49

AGPL-3.0 source for the desktop app, with the Pro build under a separate commercial licence (30). Desktop retention settings and the website privacy notice (last updated 4 November 2026) are documented. The notice says the Pro desktop app sends usage and error metadata to Supermetry with no switch to turn it off, while the open-source app's analytics stay opt-in (15). A dated notice that Tauri 0.5.3 is the final Tauri release, with a migration guide, and no stated notice period (5). Open-source desktop analytics are opt-in and disclosed in docs/electron-privacy.md (15). 30+15+5+15 = 65.

Inworld Instant Voice Cloning 55

Graded for voice cloning as the other listings in this category were, 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. Revocable workspace keys that are read-only for the Voice API until a write toggle is set, Realtime-only keys and single-use tokens, sent in a header (25 of 30). The API has no consent or verification field and no watermark was found. Only the Portal asks the user to confirm rights, and the Terms (11 June 2025), now read, have no clause on cloned voices. The write toggle keeps a default key from cloning (6 of 20). The Terms say Inworld won't train generally available models on non-public Materials except feedback and trust and safety content, while the Privacy Notice lists training among its uses. Zero Data Retention excludes cloning samples and no retention period for them is published (7 of 15, was 4). Usage is attributed per key in the dashboard and an experimental usage records endpoint exists from the Builder plan. No audit log found (7 of 15). GDPR and SOC 2 Type II are stated on every plan on the pricing page, and the security page links a vulnerability report form. No security.txt (404 per the 30 October inworld-tts check) and no bug bounty found (10 of 20, was 8). 25+6+7+7+10 = 55.

VoiceStudio 45

Local server with an optional API key (OMNIVOICE_API_KEY) and a PIN for Local network sharing. The guide warns that plain HTTP sends the PIN unencrypted and that loopback requests through a local proxy bypass the PIN (10). File reads and writes are confined to OMNIVOICE_MCP_BASE_PATH, retention deletions ask for confirmation, and since 0.5.7 the local API refuses requests from other websites and unrecognised host names and URL imports refuse private-network addresses by default (15). Transcription returns text from untrusted audio, and no prompt-injection guidance was found (0). Local generation history is visible to the operator (10). .github/SECURITY.md asks for private reports through GitHub Security Advisories or security@voicestudio.sh, with acknowledgement within 48 hours and fixes within 30 days for critical issues. A disclosure policy with a stated acknowledgement time scores 10, and whether advisories are published couldn't be read. 10+15+0+10+10 = 45.

Inworld Instant Voice Cloning 40

No x402, MPP or L402 (0). The pricing page, unread on 9 October and read on 6 November, publishes per-character rates for speech from a clone, $25 per million characters on TTS-2 and $15 on TTS-2 Flash on the free plan, and no charge for creating a clone (20, was 10). The On-Demand plan is free with 100 custom voices and instant cloning, and no card is asked for (20). inworld login opens a browser and signup is a human flow. The CLI mints keys only after that (0). 0+20+20+0 = 40.

VoiceStudio 30

No machine payment protocol (0). The Pro page lists Pro at $99 per user a year at launch ($199 list) and Lifetime at $199 one-time ($499 list), plan prices only, with purchases marked as coming soon (10). The AGPL desktop app is free to download and run with no account (20). Pro needs an account and a checkout, and the 7-day Pro trial needs an account (0). 0+10+20+0 = 30.

Inworld Instant Voice Cloning 64

The clone response echoes up to 15 seconds of processed audio as base64 with no way to leave it out. The MCP tool returns a compact result, but the server loads 40 tools with no toolsets, registering only those the credentials can call (13 of 25). Voice lists take an AIP-160 filter, pageSize, pageToken and orderBy, including source = "IVC", and a capacity endpoint reports used and maximum custom voices (20). Errors are a numeric code, a message and details, and the one example reads clearly, but no list is documented for the Voice API (12 of 20). No idempotency key, no MCP annotations in the 1.5.0 package, and the SDKs retry 5xx and network errors twice by default (4 of 20). Two required fields, language auto-detection, and official Python and Node.js SDKs (15). 13+20+12+4+15 = 64.

VoiceStudio 70

Nine tools (generate_speech, clone_voice, describe_voice, design_voice, transcribe, list_voices, list_personalities, list_languages, check_health) with docstrings of about 4,600 characters in all, read from the server source on 5 November 2026 (25, ten or fewer compact tools). Output modes keep audio out of context (OMNIVOICE_MCP_OUTPUT_MODE=files returns a URL and path), and recent history returns at most 20 items (20). A missing voice returns a JSON object with an error message, and path arguments without a base path are refused with a reason (10, the earlier partial, kept). File inputs are confined to the base path, but no tool carries readOnlyHint or destructiveHint annotations and no retry guarantee is documented (5). Client configuration examples and sensible defaults (WAV output, agent-bound voice), no official SDKs (10). 25+20+10+5+10 = 70.