WellSaid prices paid Studio usage around finished audio downloaded, not text entered or previews generated. Teams can audition voices and regenerate takes without consuming their download-minute allowance, then spend minutes only on output selected for production.
> Distinctive strength: Paid Studio allowance is tied to finished audio downloaded, allowing teams to audition and regenerate before spending production minutes. > > Where it stops being an advantage: Languages, collaboration and API access vary by plan, and unlimited preview does not guarantee an accepted performance.
That model rewards disciplined approval, but it does not make every plan equivalent. Starter and Pro are English individual plans. Business adds team workflows. Enterprise adds global languages, translation, SSO and deeper controls. The API is separately contracted and metered in characters.
Unlimited generation cannot guarantee an accepted performance; “global voices” does not establish a single sourcing model; a SOC 2 badge does not replace scope review. This review keeps those claims attributed and asks: can the exact voice and workflow reach approval efficiently, and does the chosen plan provide the rights, language, collaboration and assurance required?
Price approved downloads, not generated auditions
| Requirement | Shortlist implication |
|---|
| Low-volume commercial English | Starter: 20 min monthly or 240/year upfront, 10 projects, MP3/24 kHz. |
| Frequent individual production | Pro's 180 monthly or 2,160 annual minutes and unlimited projects offer much lower unit cost. |
| Team approval and pooled capacity | Business: 1–5 seats, pooled minutes, workspace/comments/shared pronunciation and support SLAs. |
| Non-English or translation | Enterprise is the published Studio entitlement; do not infer it from API voice lists. |
| Self-serve cloning | Exclude: WellSaid says it does not permit independent voice cloning. |
| Contracted brand/custom voice | Evaluate separate Avatar agreement, written consent, recording and commercial terms. |
| Application TTS | Evaluate separate API limits, models, character pricing and latency. |
| Sensitive content | Strong no-training/security claims; still review SOC 2 scope, retention and DPA. |
How does downloaded-minute pricing work?
Paid Studio plans allow unlimited generation and regeneration. Allowance is deducted by duration when finished audio is downloaded. A three-minute take consumes three downloaded minutes regardless of how many previews preceded it.
June 2026 introduced this meter for new plans. Annual individual customers receive the whole annual bucket upfront. Existing Business/Enterprise contracts may remain on legacy download-count terms until their contract transition; new members follow the account's current system.
Current pricing:
| Plan | Billing | Downloaded audio | Base price/minute |
|---|
| Trial | $0 | 3 min/month | — |
| Starter monthly | $19/month | 20 min/month | $0.95 |
| Starter annual | $120/year | 240 min upfront | $0.50 |
| Pro monthly | $49/month | 180 min/month | ~$0.272 |
| Pro annual | $396/year | 2,160 min upfront | ~$0.183 |
| Business | $1,920/year/user | 2,880 min/year/user | ~$0.667 |
Those figures exclude human review, rejected downloaded takes, storage/integration and add-ons. Starter, Pro and Business minutes-based accounts can buy an additional monthly or annual minute pack, one active add-on at a time, but the public help inspected does not publish stable prices. Verify the pack in account before treating it as overflow economics.
Why can Pro cost less than Starter for real production?
Starter is cheap in dollars but has 10 active projects, MP3 at 24 kHz and only 20 monthly minutes. Pro provides unlimited projects, 180 monthly minutes, MP3/TXT, up to 48 kHz and Adobe Express.
For a team producing 120 finished minutes/month:
- one Pro monthly plan includes all 120 within $49;
- Starter includes 20 and requires an add-on or workflow change;
- on annual Pro, the 2,160-minute upfront pool averages 180/month but can flex across seasons.
The correct comparison is not “$10 versus $33 monthly equivalent.” It is plan cost plus overflow plus reviewer time plus any quality/integration requirement. A low-volume creator may still prefer Starter. A recurring course/video team can hit Starter's minute and project limits immediately.
Which exports and quality levels are available?
The current comparison table lists:
- Trial/Starter: MP3, 24 kHz;
- Pro: MP3 and TXT, up to 48 kHz;
- Business: MP3, WAV, OGG and TXT, 24/48 kHz;
- Enterprise: those formats up to 96 kHz;
- paid plans: SRT/VTT caption downloads;
- Pro: Adobe Express;
- Business/Enterprise: Premiere Pro and Express.
The unusual format ladder matters. A Starter buyer who requires WAV should not assume all paid plans include it. Test the actual NLE/LMS/telephony import, loudness, sample rate, silence, caption timing and take assembly before producing a library.
Can you revise without paying twice?
Yes for generation: WellSaid says paid plans include unlimited takes. Project sections retain take history, letting users regenerate, compare, restore and delete takes before download.
Build approval around this advantage:
- lock script facts and names;
- generate several voices/styles;
- blind-review previews;
- restore the winning takes;
- run final pronunciation/timing approval;
- download only the approved assembly.
If a team downloads every candidate before review, it defeats the pricing model. Track downloaded minutes versus accepted minutes. A 90-minute course with three fully downloaded candidate casts consumes 270 minutes even if only one ships.
Deleting a project permanently deletes its audio according to current help. Download required masters before freeing a Starter project slot. This is immediate application behavior, not a complete statement about backups or processors.
Which languages are included?
Individual plans list all English voices, including multiple styles and accents. Enterprise lists all languages and translation. The API documentation includes WellSaid and partner voices across Arabic, French, German, Italian, Japanese, Korean, Portuguese, Spanish and other languages.
Do not transfer the API catalogue to an individual Studio plan. For each needed locale confirm:
- Studio/Enterprise entitlement;
- first-party WellSaid versus Partner voice;
- model compatibility (Caruso versus legacy);
- target accent/region;
- translation source and review workflow;
- pronunciation/custom dictionary support;
- native-review acceptance;
- lifecycle/replacement notice.
Caruso is currently documented as English-only. A voice row can also be legacy-only. Preserve voice ID, model and final masters to reduce migration risk.
Does WellSaid offer voice cloning?
Not as self-serve cloning. WellSaid's ethics page says it has never allowed users independently to clone voices. That is a meaningful product boundary, not a missing checkbox.
WellSaid does offer contracted Custom Voice/WellSaid Avatar work. The separate agreement requires express written consent from each person in submitted recordings. WellSaid says its stock actors provide written consent and receive residual compensation.
Choose this model if managed consent and professional voice development are more valuable than instant cloning. Exclude it if your workflow requires users to create disposable clones on demand.
For a custom voice request, obtain:
- speaker agreement and compensation;
- recording/script specification;
- permitted organization, users, channels and term;
- generated-content rights;
- disclosure/moderation requirements;
- access and audit controls;
- revocation and deletion procedure;
- model portability/termination;
- replacement and service continuity.
What does Business add for teams?
Business costs $160/month/user billed annually, supports 1–5 seats and pools 2,880 annual minutes per seat. A two-seat minimum-shaped example would be $3,840/year and 5,760 pooled minutes, though checkout and actual seat count must control.
Current product/help surfaces document:
- one Team Owner;
- Admin and Member access levels;
- team workspace/shared drives;
- pooled minutes and unlimited projects;
- collaborator seats;
- shared pronunciation library;
- commenting with notifications;
- team favorite voices;
- live chat and invoicing;
- Premiere Pro/Express integrations.
Pricing also lists access control features, but the exact permission and audit matrix needs a live workspace/contract check. Test who can invite, delete projects, download, view custom content, change billing and transfer work. Removing a member does not reduce purchased seats; the seat can be refilled.
Enterprise adds custom seats/workspaces, SSO, team analytics, custom content moderation, onboarding, dedicated success above thresholds and custom terms.
How strong are the security and data commitments?
WellSaid makes unusually concrete current claims:
- customer input/output is never used to train its models;
- customer content remains customer-owned;
- data is processed/stored in US-hosted Google Cloud and managed by US engineers;
- TLS 1.2 in transit;
- AES-256 at rest;
- SOC 2 Type I and Type II certification;
- GDPR and EU AI Act alignment;
- subprocessors and reports through a Trust Center;
- status via public status page.
These are stronger than a generic “secure” statement, but procurement should still inspect the actual current SOC 2 Type II report, scope, period, exceptions and bridge letter. Confirm whether Studio, API, partner voices, custom voices and all subprocessors are in scope.
The privacy policy uses purpose-based retention—service delivery, legal obligations, disputes and enforcement—without publishing a precise object-by-object schedule. Exact lifecycle for deleted projects, takes, imported files, API text/audio, backups and partner processors remains unknown publicly.
For sensitive work obtain DPA, subprocessors, regional requirements, deletion schedule, incident notice and least-privilege evidence before upload.
What support commitments are published?
Individual plans list email support but no response SLA. Business publishes email within one business day and live chat within four hours during business hours (weekdays 6am–4pm PST). Enterprise publishes one-business-day email and one-hour priority chat during those hours.
These are response commitments, not uptime remedies. API or mission-critical use still needs the governing SLA, maintenance notice, severity definitions and service credits in contract.
What are the API boundaries?
WellSaid API is separately contracted. Current documentation says:
- default key: 3 requests/second;
- maximum 1,000 characters/request;
- monthly character quota with billed overages rather than hard stop;
- higher concurrency by agreement;
- streaming audio response;
- voice metadata and model selection;
- pronunciation libraries;
- limited SSML plus AI Director controls for tempo, pauses, loudness, pitch and respelling.
Public stable character pricing was not established. Do not infer Studio's downloaded-minute cost. Obtain included characters, overage rate, minimum, model multipliers, failed-request billing, concurrency, regional processing and SLA.
For a controlled test, send the same 1,000-character corpus at one and three requests/second. Record response headers, time to first byte, full render time, errors and output consistency. Then repeat with difficult pronunciation, SSML and each required model/voice.
When WellSaid's downloaded-minute model is the wrong fit
Do not shortlist WellSaid when:
- you need self-serve voice cloning;
- non-English Studio output is required without Enterprise;
- Starter's 10 projects/MP3/24 kHz cannot satisfy delivery;
- you need WAV/OGG but budget only Starter/Pro;
- precise object-level retention is mandatory and contract evidence is unavailable;
- API pricing/limits are needed but only Studio has been evaluated;
- partner/legacy voice continuity cannot be accepted;
- downloaded candidates—not approved finals—would consume the minute pool.
How should you test WellSaid before buying?
Use a real 60-minute English production:
- Divide it into sections with difficult names, numbers, acronyms and delivery changes.
- Generate three voice/style candidates without downloading.
- Use take history to restore winners after script revision.
- Blind-review with target listeners; record acceptance and reviewer time.
- Download only the winning assembly; reconcile exactly 60 minutes.
- Inspect MP3/TXT or WAV/OGG, sample rate and SRT/VTT by intended plan.
- Import through the intended Adobe/NLE/LMS workflow.
- For Business, test Owner/Admin/Member, comments, pronunciation, pooled usage and offboarding.
- For non-English, validate the exact Enterprise voice/model with native reviewers.
- For API, benchmark 1,000-character requests and obtain price/SLA/security terms.
The core score is downloaded minutes per accepted minute, plus review time and governance. Unlimited previews can improve economics only when the team delays download until approval.
Does WellSaid reduce the path to an approved take?
WellSaid is a strong shortlist for English commercial voiceover teams that value unlimited iteration before paying for finished output. Pro offers unusually low base minute economics for an individual; Business trades a higher rate for pooled collaboration and support. Enterprise is the route for global languages, translation, SSO and custom assurance.
It is not a self-serve clone platform, and it should not be evaluated as one. It also leaves API price and object-level retention to contract rather than a public one-size-fits-all statement. Choose it when its exact voice passes blind review and the download-based workflow fits your approval discipline.
Main unknown: the exact retention/deletion lifecycle for project, generated and API data—including backups and partner processors—under the intended contract.
Continue with the AI Voice category, compare programmatic delivery in the AI Voice API guide, or apply the voice-generation cost protocol to accepted downloads.
Sources checked