HeyGen Voice

# Fix list: HeyGen Voice

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/heygen-voice, the October 2026 research run, assessed 10 October 2026. Grade BB, 71.3 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on HeyGen Voice: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Payments & pricing, 30 out of 100, up to 8.8 more on the total

Why it scored 30: No x402, MPP or L402 in the llms.txt, the OpenAPI file or the voice pages (0). Instant-voice speech is $15 per million characters in the public OpenAPI descriptions, and professional speech is 0.6 credits a minute. Self-serve API pricing in the dashboard needs a login, so this rests on the OpenAPI text (20). No free API allowance was found. The web Free plan and OAuth bill subscription credits, which were not checked for HeyGen Voice (0). Stripe Projects lists heygen/api, which lets an agent create the account and receive a key through the operator's Stripe login and saved payment method. A person still has to log in to Stripe once (10). Total 30.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 2. Security & auth, 69 out of 100, up to 5.4 more on the total

Why it scored 69: Model reading of the checklist, with training and retention in place of least-privilege and injection lines. API keys can be restricted to resource read and write scopes, a key without the scope gets 403 insufficient_api_key_scope, and the docs cover rotation (30). The privacy policy (11 August 2026) and the terms (23 July 2026) allow training on user input, with an opt-out by email. The security page excludes enterprise data by default (10). No retention period for voices or speech was found, deleted data stays in backups for 60 days, and voices can be deleted through the API (5). Each response carries an X-Request-Id used in usage and billing records, and the usage page has an Activity tab. Audit logs of user and admin actions are for enterprise customers (10 of 15). A valid security.txt (expires 31 December 2026), a disclosure policy with safe harbour, SOC 2 Type II stated on the security page with reports on request from security.heygen.com, and no bug bounty. The report was not read, so the certification counts as a vendor claim (14 of 20). Total 69.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 3. Reliability, 75 out of 100, up to 5 more on the total

Why it scored 75: Status page at status.heygen.com with components for www, app and api.heygen.com and dated incident history (20). It has no voice component. In the 90 days to 10 October, majors hit Video Agent (9 September, about 3.5 hours of 404s on new sessions) and Seedance (16 September, 2 minutes), and minors hit video workflows. None names speech or voices, so minor only for this surface (20). The speech endpoints publish 30 requests a minute per workspace member (15). A 429 carries Retry-After and X-RateLimit headers, the error reference asks for exponential backoff, the OpenAPI file marks 502, 503 and 504 safe to retry, and voice creation takes an Idempotency-Key (15). The terms disclaim uninterrupted service and no SLA was found (0). Instant voice creation is in a preview and the professional mode carries no label, so half (5). Total 75.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 4. Maintenance & community, 65 out of 100, up to 3.1 more on the total

Why it scored 65: Read as a model. The changelog's newest voice entry, instant voices and text-to-speech, is dated October 2026 (30). Four voice entries in the last 90 days, professional voice clone, pause tags and ElevenLabs model selection in September and instant voices in October, and dated sunset notices elsewhere in the API, v1 and v2 on 31 October 2026, announced in April (20). Public changelog with RSS and a developer Discord. Answers were not sampled (10 of 15). No official SDK. HeyGen ships a CLI and agent skills (5 of 15). No package health record was read (0). Total 65.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## 5. Agent ergonomics, 83 out of 100, up to 2.8 more on the total

Why it scored 83: API reading of the checklist. The completed response is a URL and a duration, streaming returns WAV parts as they are made, and timestamps come only when asked. Output is WAV only, with no format or sample-rate choice (20 of 25). Voices list newest first with `limit` and `token`, and text is capped per request, but there are no output-size controls beyond the text (15 of 20). Typed error codes such as voice_not_ready, voice_expired and insufficient_credit, each with a documented next step (20). Idempotency-Key on voice creation, safe-retry status codes on speech, failed or disconnected requests not charged, and a best-effort `seed` for professional voices (20). Only `voice_id` and `text` are required and `model` defaults to heygen-voice-1, but no official SDK was found, only the HeyGen CLI and agent skills (8 of 15). Total 83.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 6. Transparency & trust, 69 out of 100, up to 2.7 more on the total

Made of editorial 52, provenance 85.

Why it scored 69: Closed service with terms dated 23 July 2026. Paid plans own their output, and free-plan output is limited to non-commercial use (15). Privacy policy, terms and security page agree that non-enterprise input can train models with an opt-out by email, and differ on the entity name (HeyGen Technology Inc. and HeyGen, Inc.). No retention period for voice or speech data (12 of 30). Dated deprecation notices in the changelog, such as the v1 and v2 sunset on 31 October 2026, and no notice yet for this model (15 of 20). Subprocessors are listed on the security portal, which was not read, and the privacy policy says data may be processed in the United States (10 of 20). Total 52.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Domain age: heygen.com, no registry record we could read (0 of 15)

## 7. Schema & documentation, 94 out of 100, up to 1 more on the total

Why it scored 94: OpenAPI 3.1 file at developers.heygen.com/openapi/external-api.json covering create, list, get and delete voice, completed speech and streamed speech (25). llms.txt with a Markdown twin of every page (10). Descriptions say what each endpoint is for and when to use another, for example that Turbo refuses professional voices and that catalogue voices go through /v3/voices/speech. The model page and changelog do not mention Turbo (17 of 20). Request bodies are typed with an enum for `model`, numeric ranges for every control, a 5,000-character cap and `additionalProperties: false` (15). curl examples for both endpoints and a per-endpoint error table with codes and recovery steps (15). Versioned /v3 paths and a public changelog with an RSS feed, dated by month only (12 of 15). Total 94.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- unchecked: the self-serve API pricing modal at app.heygen.com/developers/api, which needs a login. The dollar value of an API credit is inferred from the OpenAPI example of 9/60 credit for $0.15.
- unchecked: whether the web Free plan or OAuth credits can call the HeyGen Voice speech endpoints.
- unchecked: the subprocessor list and SOC 2 report on security.heygen.com.
- The claim that HeyGen Voice leads a text-to-speech arena came from a social media post. No first-hand page with the ranking was read, and the listing does not state it.
- No release day for either mode. The changelog dates them by month only.
- No authenticated call was made, so latency, the Turbo model and the streaming format are as documented, not measured.

## Weaknesses

- Speaks only voices cloned in the caller's workspace. Stock and designed voices go through a separate endpoint and engines
- 30 requests a minute per workspace member on the speech endpoints, on every plan
- Output is a 44.1 kHz mono WAV only, and text is capped at 5,000 characters a request
- HeyGen's privacy policy allows training on user input, with an opt-out by email. Enterprise data is excluded by default
- No official SDK, no free API allowance and no SLA in the terms

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Create a voice first with POST /v3/models/audio/voices and `mode`, then poll GET /v3/models/audio/voices/{voice_id} until `status` is ACTIVE before calling speech
- Send `expressiveness_boost` only for instant voices and `seed`, `speed`, `pitch_shift`, `pitch_variance` or `<break>` tags only for professional voices. Mixing them returns 400 invalid_parameter
- On the stream, decode and play each base64 WAV part in `part_index` order. Do not concatenate the bytes
- Send an Idempotency-Key when creating a voice, and retry 502, 503 and 504 on speech, which the OpenAPI file marks as safe
- Use POST /v3/voices/speech for stock or designed voices. It is a different engine and price

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.