AI COMPANIES / RESEARCH DOSSIER

WellSaid.

Inside the company behind enterprise AI voice: founders, proprietary TTS, licensed Voice Avatars, funding, customer evidence, pricing, security and market position.

Company profileGenerative voice AI2026 editionEvidence-led

Last reviewed 04 November 2026 · Primary sources first · Publisher-ready editorial format

WellSaid logo supplied by the publisher
Original supplied logo · unmodified
THE EXECUTIVE VIEW

WellSaid is not simply an AI voice generator. Its operating story is an enterprise production system built around licensed actor voices, content controls, collaboration and programmable synthesis—while its legal, funding and consent record requires careful distinctions.

WellSaid at a glance

LEGAL ENTITYWellSaid Labs, Inc.“WellSaid” is the current brand.
BASED INSeattle, WashingtonPrivate U.S. company.
PRIMARY MARKETEnterprise Voice AIProfessional content production + API.
FOUNDING TIMELINE2018 / 2019Project origin in 2018; current About page says founded 2019.
PUBLIC EQUITY ROUNDS$12M$2M seed and $10M Series A; separate 2026 debt amount undisclosed.
CURRENT CEOBenjamin DorrVerified on company leadership page.

WellSaid is a U.S.-based, privately held AI voice company making synthetic speech for training, learning, marketing, and products. Its main offerings combine a browser-based production studio, organizational workspaces, consent-based voice creation, and a text-to-speech API. Rather than replacing all human performance, the platform focuses on repeatable licensed narration and rapid revisions.[S1] [S4] [S7]

Evidence note
The figures on this page refer to different things. $12M covers two publicly reported equity rounds; venture debt closed in March 2026 but the amount was not disclosed. The founding date is disputed across current sources. The profile does not manufacture a combined total or a current valuation.

About the company · brand, legal entity & contacts

The public-facing name is WellSaid. Its 2025 website Terms, Privacy Policy and 2025 service agreements identify the contracting corporation as WellSaid Labs, Inc.. A name change of that legal company has not been substantiated. Historical coverage and investor pages often use “WellSaid Labs,” while the current company website generally uses “WellSaid.”[S1] [S2] [S3] [S13]

FieldVerification
Brand / operating nameWellSaid
Legal operatorWellSaid Labs, Inc. (official terms)
Country / HQUnited States · Seattle, Washington; registered correspondence address at 113 Cherry St, Ste 52562, Seattle, WA 98104-2205
FoundedOfficial About: 2019; historical 2018 AI2 research origins. Incorporation date not independently established.
Status / listingOperating; privately held (LinkedIn). No public ticker identified.
Parent / subsidiariesNo parent or subsidiary verified in reviewed primary materials. Cap table not publicly disclosed.
Websitewellsaid.io
Supportsupport@wellsaidlabs.com
Voice talentactors@wellsaidlabs.com
Security inquiriessecurity@wellsaidlabs.com (historically published); current Trust Center available
Developer documentationdocs.wellsaidlabs.com
Official company LinkedInlinkedin.com/company/wellsaidlabs
YouTube / X / GitHubNo current verified corporate account URL included without an authoritative matching reference.
Careers / pressCompany About and Leadership pages link to careers and press inquiry channels.

The founding-year discrepancy should stay visible: the current About page says 2019, while LinkedIn lists 2018 and 2021 reporting locates the earlier work inside the Allen Institute for AI incubation effort. These may reflect research inception versus company formation; the precise incorporation date has not been verified.[S1] [S31] [S28] [S29]

Founders & verified current leadership

Two co-founders appear in independent 2021 reports: Matt Hocking and Michael / Mixail Petrochuk. The official leadership page currently spells the technical founder’s first name Mixail; historical coverage commonly spells it Michael. These are treated as naming variants for the same technical co-founder in the reviewed sources.[S28] [S30] [S4]

PersonPosition / statusBackground / contributionEvidence
Matt HockingCo-founder; historical CEOCreative and product experience in agencies and startups; helped translate AI voice research into corporate content products. Not listed as the current CEO.[S28] [S30]
Mixail (Michael) PetrochukCo-founder; CTOUniversity of Washington computer-science degree; PyTorch-NLP and applied deep-learning background. Leads voice research and proprietary TTS technology.[S4] [S30]
Benjamin DorrCurrent CEOFormer executive at Cordial, Valassis and Rhythm NewMedia; investment background at Carlyle Group and banks.[S4]
Chris JohnsonChief Product & Technology OfficerFormer Microsoft leader and Hyperfish co-founder; oversees product and engineering.[S4]
Natalia HoltsSVP, Finance & AccountingFormer Sensor Tower and Cordial finance executive; leads finance operations.[S4]
Elise KubickiVP, Customer ExperiencePreviously at Gorgias and Stitch Fix; customer onboarding and expansion.[S4]
Billy ShengVP, Business DevelopmentExperience at Esper and AppSheet; enterprise partnerships and sales.[S4]
Megan GodwinVP, PeoplePreviously at Code.org; people and organizational development.[S4]
Jeni BishopVP, MarketingPreviously at Cordial and other SaaS organizations; brand and go-to-market.[S4]
Leadership clarification
The official leadership page currently lists both Chris Johnson (Chief Product & Technology Officer) and Mixail Petrochuk (Chief Technology Officer). This profile preserves those exact published titles instead of inferring a reporting hierarchy.

The current involvement of Matt Hocking after his historical CEO tenure is not established by the reviewed current company leadership page. No unsupported founder shareholding or board-seat information is assigned.[S4] [S28]

Founding story & company timeline

The early WellSaid effort grew out of the Allen Institute for AI ecosystem in Seattle. The founders aimed to make synthetic voice substantially more lifelike and practical for repeatable commercial narration. Corporate training was an early target because recordings often need changes, yet actor booking, new takes and editing can slow a release. Its product later developed toward marketing, long-form content, apps and governed enterprise voice operations.[S28] [S29] [S30]

2018

Research roots at AI2

Historical reporting traces the technology to a research/incubation project in Seattle.

2019

Company / seed phase

Current About page says founded in 2019; GeekWire reports approximately $2M seed fundraising.

2021 · July

Series A investment

Oversubscribed $10M financing led by FUSE; expanded products and go-to-market.

2023 · June

Consent-based NPR collaboration

A voice model of former Planet Money host Robert Smith created with explicit permission for a limited editorial experiment.

2024–25

Collaboration and model controls

Cues/AI Director, expanded workspaces, pronunciation tools; Caruso became available in Studio and API in March 2025.

2026 · March

Enterprise localization + financing

Enterprise Studio translation launched; venture debt financing from Multiplier Capital announced.

2026 · June

Studio/pricing transition

Starter and Pro plans and downloadable-minute accounting introduced; legacy Studio project migrations began.

2026 · Aug–Sep

New generation workflow tools

Generate All, Audition voice matching, voice model previews and expanded Insights Dashboard CSV export.

Dates reflect when an event was reported or a feature appeared in official release notes, not a guarantee that the feature was available to every account on that date. “Preview 13.2” is an upcoming-model preview, not a confirmed released model family.[S9] [S6] [S41] [S27]

Ownership, capitalization & funding

WellSaid is venture-backed and described as privately held. Its known investors include FUSE, Voyager Capital, Qualcomm Ventures and GoodFriends. AI2 incubator participation is documented as an early origin, but that alone does not prove an ongoing board seat or equity position. No reviewed public source supplies an authoritative current cap table.[S28] [S29] [S30] [S31]

DateFinancingDisclosed amountPartiesEvidence
2019Seed equityApprox. $2 millionVoyager Capital reported among early investors; round specifics incomplete[S29]
7 July 2021Series A equity$10 millionFUSE (lead); Voyager, Qualcomm Ventures, GoodFriends[S28] [S48]
17 March 2026Venture debtAmount not publicly disclosedMultiplier Capital[S27]
DISCLOSED EQUITY$12 millionSum of two independently reported equity rounds; not inclusive of undisclosed venture debt.
CURRENT VALUATIONUndisclosedNo current company-announced valuation established.
REVENUE / ARRUndisclosedIndependent audited revenue and profitability data were not found.

Dealroom displays a historical 2021 valuation figure, but it is third-party database information rather than an independently validated company valuation. It should not be described as current. Some funding aggregators omit the 2019 seed or the 2026 debt; this profile keeps the instruments separate. Ownership percentages, voting control, subsidiaries, and investor board assignments: not publicly disclosed in the reviewed sources.[S47] [S27] [S29]

The corporate model is primarily paid software access: individual subscriptions, team seats, customized enterprise contracts, API access, and negotiated voice-related services. Standalone custom-voice revenue and professional-services billing are possible offerings, but precise revenue mix and deal terms are not publicly disclosed.[S5] [S42] [S3]

Current products & commercial positioning

WellSaid is best described as a hybrid enterprise AI voice production platform: Studio makes repeatable business narration, team features manage the process, custom voices support approved sonic identities, and an API embeds speech inside software and media pipelines. It is more than a basic text-to-speech converter, although it remains specialized in speech generation rather than a general conversational-assistant stack.[S1] [S7] [S43]

01 / CORE SAAS

WellSaid Studio

Script-to-speech production, voice and style selection, AI Director controls, clip editing, export and captions. Input: text; output: recorded speech assets.

Individual · Business · Enterprise
02 / LIBRARY

WellSaid Voices

Licensed actor-based Voice Avatars, styles and regional voice options. Preview and select appropriate voices per project.

English on self-serve · global languages on Enterprise
03 / GOVERNANCE

Team Workspaces

Shared projects, collaborators, team voice favorites, pronunciation dictionary, commenting and administrative controls.

Business · Enterprise, tier dependent
04 / DEVELOPMENT

WellSaid API

Authenticated text-to-speech generation for applications, batch workflows and streaming audio. Technical users send text and receive synthesized audio.

Separately sold API plan / sales-led access
05 / SPECIAL PROJECTS

Custom Voice

Authorized, actor-consented voice models for a unique brand or organizational need. A production and licensing workflow, not an unrestricted upload-and-copy tool.

Availability and pricing via company engagement
06 / CONNECTORS

Adobe Integrations

Adobe Express voiceover add-on and Premiere Pro workflow access, according to product pricing and integration guides.

Plan-dependent
Product evolution
Older materials refer to legacy Studio, Standard, Caruso (Legacy), Voice Avatars and other historical project labels. June 2026 migration guidance says these labels can identify old projects. Do not assume a legacy project type is a currently sold product.

WellSaid describes its voice catalog as “280+” in a current marketing-use-case page while another official page’s headline refers to “240+.” Both are company-reported marketing counts; the exact live production-ready total has not been independently reconciled.[S17] [S44]

How WellSaid works

01ScriptWrite or paste text
→
02Choose voiceAvatar + style
→
03DirectPronunciation + cues
→
04ExportAudio / captions

Studio workflow: An instructional designer writes a script, selects a licensed voice, generates takes, corrects names and abbreviations, adjusts pitch or pacing when supported, reviews sections with teammates and exports final audio for a course. Small copy changes are handled by regenerating affected clips instead of scheduling a new studio recording.[S5] [S9] [S43]

Enterprise workflow: Admins organize teams and project workspaces, designate preferred voices, maintain reusable pronunciation rules, collect comments and control access. Higher tiers can add SSO, usage reporting and procurement controls. Formal approval sign-off with audit trails should not be assumed without checking the specific enterprise agreement.[S5] [S3]

AYour appText / selected voice
→
BWellSaid APIAuthenticated REST request
→
CSynthesized speechStream / audio file
→
DDeliveryApp / video / service

The API supports both low-latency streaming and asynchronous workloads, according to the current product page. Streaming TTS is not the same thing as a complete conversational AI agent; developers still need their own speech recognition, dialogue, orchestration and safety layers when building a live assistant.[S7]

AI text-to-speech technology, models & training data

WellSaid says its synthesis stack is proprietary, patented or patent-protected and based on private licensed voice datasets. Historical accounts describe neural speech-generation research and a transition from early Tacotron experimentation to in-house long-form synthesis. Those descriptions do not establish the architecture of its current production models. WellSaid does not publicly document a full model-layer diagram, parameter counts, network weights or architecture for Caruso.[S11] [S10] [S28] [S27]

Model / labelWhat is verifiedAvailability / caveat
legacyOriginal API speech engine; default when API model field omittedCurrently listed in official API docs [S8]
carusoAdvanced model supporting AI Director pitch, tempo, loudness and respelling cuesStudio and API; API guide says English-language voices only [S8] [S9]
Preview 13.2An upcoming model available for evaluation in the preview selectorPreview, not confirmed general release as of September 2026 [S9]
Internal model weights / third-party model providerArchitecture and training-code implementation specifics not publicly disclosedDo not assert use of OpenAI, Azure TTS or any open-source foundation model
General TTS explanation — not a claimed WellSaid architecture
In neural text-to-speech, a system analyzes text and pronunciation, predicts timing and expressive acoustic properties, and synthesizes an audio waveform. Some systems use phoneme models and neural vocoders; others use newer audio-token generators. WellSaid has not publicly documented every one of these stages for its present models.

The company says its training data includes licensed recordings from professional actors who gave written permission, along with licensed or appropriately permitted scripts. The actor program says individuals are paid for recording sessions and usage of their synthetic voice. Exact dataset sizes, speaker counts, recording hours, language balance, preprocessing pipeline, and any synthetic-data mix: not publicly disclosed.[S10] [S11] [S14]

The 2026 DPA names Google Cloud for storage/TTS processing and Azure for language mechanics. A subprocessor relationship does not establish that WellSaid licenses another company’s core TTS model. The company describes its voice model as proprietary.[S12] [S11]

Voice Avatar vs custom voice vs voice clone

TermWhat it means in this contextRight / access implications
Standard / library TTS voiceA ready-to-use synthetic narrator that reads arbitrary text.Subject to the chosen plan’s licensed commercial usage terms.
WellSaid Voice AvatarWellSaid’s brand term for a modeled voice built using a consenting professional performer’s recordings.The library voice is not automatically owned by the customer. [S10] [S14]
Custom Brand VoiceA specific synthetic voice commissioned or arranged for an organization and its approved use case.Talent consent, commercial usage scope and contract govern rights. [S3] [S23]
Voice CloningTechnical process of modeling a recognizable human voice from authorized recordings.WellSaid agreements contemplate consent-based custom voice models; they require express written permission. [S3]

Is cloning offered? The current Business/Enterprise services agreement explicitly discusses a feature the company may make available to create a “Custom Voice Model” from authorized audio. The ethics page separately says ordinary users have never been permitted to independently clone any arbitrary voice. These statements are consistent if custom voice cloning is a permissioned, controlled service, rather than unrestricted self-service cloning. Exact onboarding and current plan eligibility require sales confirmation.[S3] [S10]

An illustrative historical project is NPR Planet Money’s 2023 experiment with the voice of Robert Smith. WellSaid reports obtaining explicit consent, restricting the project and promising to delete the synthetic voice after the series. This was a bespoke editorial collaboration, not proof that unverified users can clone other people at will.[S41]

Licensing distinction
An organization may own the audio output under applicable contractual terms without owning WellSaid’s underlying model, an actor’s voice identity or the right to reuse a custom model outside its license. Always review the relevant services agreement and the talent contract.

Voice controls, pronunciation, languages & audio

WellSaid’s product pages describe voice styles, project-level editing, phonetic respelling, pronunciation libraries and AI Director cues for pitch, tempo, loudness and certain delivery characteristics. Shared pronunciation and approved-voice features are especially useful for company names, acronyms, medical terminology and consistent multi-author training. Exact controls can depend on the voice model and tier.[S5] [S7] [S8]

CapabilityConfirmed support / caveat
Direct controlPitch, tempo and loudness via AI Director on Caruso; pronunciation respelling and selected style/tone options [S8] [S5]
Voice stylesNarration, Conversational, Promo and Character, depending on voice [S15]
PronunciationPersonal/team pronunciation dictionaries and custom phonetic guidance [S5] [S21]
LanguagesEnglish accents on self-serve plans; non-English global languages predominantly via Enterprise [S15]
Global examplesArabic, Mandarin/Cantonese, French, German, Spanish, Japanese, Korean, Portuguese, Italian, Dutch, Persian and others in current Help Center [S15]
Voice catalog“280+ voices” marketing claim vs “240+” in another official page; counts not reconciled [S17] [S44]
Same actor across languagesHelp Center says each voice is trained in a specific language; do not assume a given voice can speak every language [S15]
Studio audio formatsStarter/Pro: MP3; Business/Enterprise: MP3, OGG, TXT, WAV. Paid plans offer SRT/VTT captions [S5]
Audio sample ratesUp to 24 kHz Starter, 48 kHz Pro, 96 kHz Enterprise according to plan list [S5]
Language-count caution
The voice catalog intro says “30+ languages,” while the Help Center’s current language-variety list names 17 non-English language categories plus English, with many regional variants. Because the catalog and marketing counts are not identical, this page avoids publishing an exact independently established language count.

Enterprise collaboration, API & deployment

FeaturePractical purposeAvailability / status
Team projects / workspacesOrganize scripts, audio and team accessBusiness and Enterprise [S5]
Comments / collaborator seatsReview and coordinate revisionsBusiness and above [S5]
Shared pronunciation / preferred voicesKeep brand and product language consistentTeam features, check plan [S5]
SSO / security controlsIdentity management and managed accessEnterprise [S5]
Usage insightsReporting and historical usage exportsEnterprise dashboard updated Sep 2026 [S9]
Custom content moderationEnterprise policy and permitted-user controlEnterprise [S5]
REST TTS APIGenerate audio from text inside applicationsAPI product [S7]
Streamed audioSupport responsive live speech outputOfficial API offers streaming [S7]
Asynchronous synthesisRun background and large-batch workloadsOfficial API lists async workflows [S7]
API authenticationAPI key auth; example uses X-API-KEYOfficial site example [S7]
Model selectionPass legacy or caruso to TTS APIOfficial docs [S8]
Official Python/Node SDKUnconfirmed in reviewed documentationDo not label third-party wrappers official
Cloud SaaS / APIBrowser Studio and hosted APIVerified product delivery [S7]
On-premises / private VPCNot publicly establishedAsk enterprise sales before assuming availability

The publicly illustrated streaming endpoint is POST https://api.wellsaidlabs.com/v1/tts/stream. It is an official example, not a complete integration contract. Production teams should consult the live API documentation for request schemas, account permissions, concurrency, error handling and plan-specific quotas; exact rate limits and all supported API encoding formats were not independently verified for this profile.[S7] [S8]

The current API messaging expressly covers real-time TTS, high concurrency and background synthesis. That capability should not be confused with a fully bundled, turn-taking call-center voice agent or an independent speech-to-text/transcription service.[S7]

Who uses WellSaid · people, industries & practical workflows

LEARNING & DEVELOPMENT

Instructional design

L&D specialist → changing compliance modules → Studio → consistent narration and captions → faster retraining without new recordings.

CORPORATE COMMUNICATIONS

Employee enablement

Internal ops team → frequent product changes → shared voice + pronunciation rules → updated walkthroughs → less narrator dependency.

MARKETING

Campaign production

Brand team → video script iterations → Studio / approved voice → voiceover variations → cleaner production handoffs.

SOFTWARE PRODUCT

Application narration

Developer → text-driven user journeys → TTS API → playable speech → embedded guided experience.

LOCALIZATION

Global training

Regional learning team → translated scripts → enterprise language voices → multilingual course audio → localized employee learning.

PRODUCTION AGENCIES

Client video work

Creative producer → last-minute copy changes → generate placeholder/final voice takes → synchronized edits → reduced re-record loops.

Named customer stories support enterprise software, aviation training, internet registry education, retail/promotional products, automotive lubricants, wellness applications, marketing production and instructional-services workflows. Healthcare is documented through XpertPatient and an anonymized large healthcare-training organization; anonymized cases must not be assigned a brand name.[S18] [S19] [S20] [S22] [S23] [S51]

Verified customers, case studies & reported outcomes

The table below lists organizations with named, readable customer stories, not merely logos. These case studies are published by WellSaid and often quote customer employees, but outcomes remain customer-/vendor-reported unless separately tested.[S18]

OrganizationProductSpecific workflowReported outcome
FlightSafety InternationalStudio · employee L&DCompliance, safety and instructor trainingRe-record cycle reduced from as long as two weeks to hours; five-person team supporting ~5,000 employees [S19]
Microsoft (internal enablement team)StudioNarration in Rise, Storyline and Camtasia; global trainingA snippet that could take 2–3 hours to self-record can now be generated in minutes [S20]
ARINStudioTechnical vocabulary, microlearning, explainersVideo update turnaround described as weeks to days [S21]
MotulStudioMultilingual global training and PowerPoint narrationNearly 80 multilingual modules produced in six months [S22]
TRIPPAPI + Custom VoiceVoice content for meditation and wellness appCustomer reports production shrank from 8 weeks to minutes; two custom voices [S23]
PROVOKE SolutionsStudioProduct/demo video scripting and timingCustomer reports 25% reduction in production time; noted that client-branded human voice may be used in final [S24]
PowerTrainStudioTraining-course narration and changesDescribed line-level revision from hours to minutes; no precise percentage published [S25]
4imprintStudioRemote-training / e-learning productionMore consistent production and easier changes; no numeric ROI established [S26]
≈80Motul modules over six months
25%PROVOKE production-time reduction
8 weeks → minutesTRIPP production timeline
2 weeks → hoursFlightSafety revision cycle

WellSaid’s own LinkedIn company description and March 2026 financing release also name Microsoft, T-Mobile, Progressive, Humana and ServiceNow among enterprise customers. A named case study for Microsoft is reviewed above; the others are company-identified customers, but exact WellSaid deployment details and independently validated impact for each were not found in the reviewed materials.[S31] [S27]

The security page states “more than 40,000 customers globally,” but does not show a methodology or break out paid enterprises, trial accounts or historical versus active relationships. Treat that as WellSaid-reported; not independently verified, not a verified count of enterprise contracts.[S11]

Case-study integrity rule
TRIPP’s “3 million downloads” describes the TRIPP app, not WellSaid adoption. Microsoft’s study is about a specific internal enablement team, not proof of universal organization-wide rollout. PROVOKE reports WellSaid’s use during production but notes its client could use a different voice for the finished video.

Current pricing · checked 4 November 2026

The live pricing page shows individual monthly and annual options, yearly Business seats and negotiated Enterprise plans. In June 2026 the company shifted to a downloaded-minutes usage measure: iterations on paid plans do not count against the finished-audio allocation; exported minutes do. Prices below are U.S. dollars before applicable taxes and refer to new customers.[S5] [S6]

PlanPublic priceIncluded finished audioImportant access / restrictions
Free / Trial$03 downloadable minutes per monthNo commercial rights; limited generation and projects; try before purchase
Starter$10/month annually ($120/year) OR $19 month-to-month240 min/year annual; 20 min/month monthly10 projects; English voices; commercial rights; MP3; 24 kHz
Pro$33/month annually ($396/year) OR $49 month-to-month2,160 min/year annual; 180 min/month monthlyUnlimited projects; Adobe Express; up to 48 kHz
Business$160/month/user, billed annually ($1,920/year/user)2,880 min/year/userUp to 5 seats; shared workspace; commenting; Premiere Pro; invoicing
EnterpriseCustom / contact salesCustom negotiatedAll global languages, translation, SSO, security/admin controls, custom workspace, up to 96 kHz
APIContact sales / API packageContract / plan dependentOfficial 7-day API trial; check usage and access terms
Custom Brand VoiceNot publicly pricedNegotiatedTalent/licensing/voice-model arrangements vary by contract

Paid plans list unlimited generation, but not unlimited final downloads. Annual minute allocations are provided yearly and unused minutes do not roll over. The company says canceled subscriptions do not provide project access while inactive and projects may be stored for up to five years after cancellation. Verify the latest terms before purchase.[S5]

Commercial use note
The $0 trial explicitly excludes commercial rights in the current plan table. Creator and paid tiers list commercial rights subject to their agreements. Commercial authorization does not override third-party copyright, performer permissions or content restrictions.

Security, privacy & customer data

WellSaid publicly states it has a SOC 2 Type II compliance/reporting program and makes a trust report available to qualifying customers on request. Its security page also describes TLS 1.2 in transit, AES-256 encryption at rest, Google Cloud-based U.S. hosting, content moderation, and support for enterprise SSO. These are company claims backed by published policy pages; the independent audit report was not personally inspected for this profile.[S11] [S10] [S5]

TopicDocumented positionInterpretation / remaining question
Training on customer dataBusiness & Enterprise agreement says WellSaid will not use Customer Data to train its or third-party AI modelsContractual policy-confirmed for covered contracts; check your own online or negotiated agreement [S3]
Licensed voice trainingActor recordings used by permission; residuals reportedly paidCompany-reported actor policy [S10] [S14]
Model separationCustom Voice Models are not shared with other customersExpressly stated in enterprise services agreement [S3]
Data hostingWellSaid says customer data processed/stored in U.S. Google CloudOther contractual subprocessors include Azure, MongoDB, etc. [S11] [S12]
Privacy rightsPrivacy policy describes personal information categories, uses and privacy requestsPolicy last updated July 2024; DPA updated in 2026 [S13] [S12]
Security assuranceSOC 2 Type II report available upon requestReport coverage and exceptions not independently reviewed [S11]
EncryptionTLS 1.2 in transit / AES-256 at rest claimedVendor-published; independent configuration not audited here [S11]
Content controlsModeration, user blocking and anti-abuse policiesExact review thresholds are not publicly documented [S10] [S11]
Retention / deletionPricing FAQ says inactive projects may be stored up to 5 years after cancellationNot a universal retention period for all script, logging, custom voice or enterprise data [S5]
Private deploymentOn-prem/VPC not verifiedDo not presume regulated private infrastructure available

The 2026 DPA identifies service providers for cloud storage, TTS processing, language mechanics, authentication/verification, moderation, analytics and API infrastructure. This is a contractual processing list, not a list of WellSaid model licensors. Teams processing regulated or identifiable data should obtain the current DPA, subprocessor register, data-residency terms and relevant security report directly.[S12]

Voice talent, consent & responsible AI

Consent and voice rights are central differentiators in WellSaid’s stated operating approach. Its published actor program describes actors being paid for recording sessions and receiving usage-linked royalties for their Voice Avatars. The ethics page says training uses private consented recordings rather than scraped open-source voice data. These are specific company policy statements, not independent third-party audits of every recording or payout.[S10] [S14]

PERMISSION

Written voice consent

Company agreements require express written authority for custom voice training and generation. Customer remains responsible for necessary rights.

COMPENSATION

Talent royalties

The actor program describes recording fees plus royalties linked to synthetic voice rendering; exact percentages are contractual, not public.

MISUSE

Content moderation

WellSaid says it moderates harmful misuse and restricts public impersonation / unauthorized cloning.

Synthetic-audio labeling / watermarking: No sufficient current primary evidence was identified here to assert a universal mandatory synthetic-voice label, robust audio watermark, or interoperable provenance standard for all WellSaid exports. Users should establish suitable disclosure practices for their context rather than assuming the vendor automatically labels every audio file.[S10] [S49]

Voice cloning nuance: WellSaid’s marketing ethics page says ordinary customers cannot independently clone any voice, while legal agreements anticipate controlled Custom Voice Models based on rights-cleared recordings. The practical distinction is authorization, access control and licensing—not proof that no cloning technique is involved.[S10] [S3]

For an enterprise procurement review, request written confirmation of actor release coverage, geographic and media licensing, duration, deletion/termination obligations, attribution (if applicable), moderation escalation, content disclosure and limits on derivative or public-figure impersonation.[S3] [S14]

Partnerships, product integrations & programs

Organization / programClassificationDocumented relationship
AI2 / Allen Institute for AI incubatorHistorical ecosystemResearch/incubation roots; not a claim of current ownership [S28] [S29]
FUSE, Voyager, Qualcomm Ventures, GoodFriendsInvestorsSeries A investors; not customers or voice technology suppliers [S28]
Multiplier Capital2026 lenderVenture-debt counterparty; debt amount not disclosed [S27]
Adobe ExpressNative add-onCreate voiceovers in Express; Starter+ described [S40]
Adobe Premiere ProIntegrationAvailable in Business plan listing [S5]
Google Cloud / Azure / othersSubprocessorsOperational data-processing roles, not necessarily product integration [S12]
Articulate Rise, Storyline, CamtasiaCustomer workflow / exportMicrosoft case study imports WellSaid audio to these tools; native integration not established [S20]
LMS / video production systemsExport / APICommon workflow via audio exports or custom API work; do not label every LMS a native WellSaid integration

Current official SDKs for Python, Node.js or JavaScript were not conclusively verified in the reviewed developer material. An API with documentation is not proof of an officially maintained language-specific SDK. Likewise, no authoritative current evidence was identified for a separate on-premises edition, private VPC product or broad OEM reseller program.[S7] [S8]

Independent user sentiment · what reviewers say

G2’s 2026 summaries show repeated praise for a straightforward production interface, realistic voice quality, variety and time savings. Repeated criticisms involve unusual names, pronunciation mistakes, editing cadence, certain voices sounding artificial, cost and account/contract frustrations. Capterra and G2 ratings are snapshots of reviewers, not controlled acoustic benchmarks; several individual reviews are older and may describe legacy pricing or Studio features.[S32] [S33]

Common positivesCommon complaints
Fast text-to-voice generation for trainingProper nouns and brand terminology sometimes require manual corrections
Natural, professional-sounding narration for many use casesSome scripts sound less natural across long passages or specific voices
Easy repeat revisions and voice experimentationFine-grained expressiveness may need several takes
Pronunciation and team consistency toolsPaid subscriptions/usage caps may be expensive for casual creators
L&D workflow fitSome user-reported support and refund dissatisfaction

A June 2026 instructional-design Reddit discussion also includes a user who described WellSaid as expensive and noted changes to the interface and model. This is anecdotal context, not an independently measured overall satisfaction rate.[S32]

Evidence quality
Marketing performance percentages, site customer totals, reviews and vendor comparison charts are not interchangeable with audited business metrics or blinded speech-quality tests. This report keeps those evidence types separate.

Where WellSaid fits in the AI voice market

The most meaningful competition falls into four groups: enterprise content studios (Murf), generative speech and custom voices (ElevenLabs), voice generation plus forensic detection (Resemble AI), and developer/cloud TTS infrastructure (Microsoft Azure Speech, Google Cloud TTS, Amazon Polly). They overlap on speech synthesis, but solve different operational problems.[S34] [S35] [S36] [S37] [S38]

PlatformCenter of gravityStudio / team workflowAPI / streamingDifferentiator / caveat
WellSaidLicensed enterprise narrationYes, strong focusYes / streamingConsent-led actor catalog and enterprise production controls [S5] [S7]
ElevenLabsExpressive multilingual voice + agentsYes / enterprise offeringsYes / low-latencyBroad speech model portfolio, cloning and real-time agents [S34]
MurfBusiness voiceover studioYes / workspace tiersYes / streaming APIStrong creator-to-enterprise workflow + developer options [S36] [S50]
Resemble AICustom voice / deepfake detectionUse-case dependentAPI-basedDetect product includes deepfake verification; private/on-prem detection options [S35]
Microsoft Azure AI SpeechCloud speech + custom voicesDeveloper portalYes / real-time & batchCloud infrastructure, developer toolchain, approval-based custom voices [S37]
Amazon PollyCloud generative and neural TTSAWS consoleYes / bidirectional streamAWS pay-for-usage infrastructure; Oct 2026 generative model upgrade [S38]
Google Cloud TTSCloud text-to-speech infrastructureCloud consoleAPIDeveloper-oriented cloud service; assess current models at purchase time
PlayHT / PlayAIHistorical streaming TTS vendorClosedClosedNot a live competitor: third-party migration documentation reports 2025 shutdown [S39]
WELLSAID vs ELEVENLABS

Managed narration vs broad voice AI

WellSaid emphasizes actor rights, L&D workflows, workspaces and control; ElevenLabs offers a larger family of expressive and low-latency models, cloning and conversational tools. Select based on language, rights, quality testing and total cost—not a universal “best.”

WELLSAID vs RESEMBLE

Production vs production + verification

WellSaid is strongest in the content-production story. Resemble AI’s public offering includes deepfake detection for audio, image and video, so it is a better functional comparison when authentication is a requirement.

WELLSAID vs MURF

Overlapping studio competitors

Both offer speech generation, collaboration and APIs. Compare actual seat permissions, editorial workflow, export rights, languages, API allowances and quoted enterprise contracts rather than vendor-authored price-per-minute claims.

WELLSAID vs CLOUD TTS

Production workspace vs infrastructure

Azure and Polly emphasize programmable speech services with extensive developer ecosystems. WellSaid packages voices, pronunciation, review and creative production for business teams that do not want to build the whole workflow themselves.

PlayHT should not be presented as an active 2026 alternative: an Inworld migration document reports the platform shut down by the end of 2025 after a team acquisition. That detail is reported by an alternative vendor rather than a direct PlayHT service notice and is labeled accordingly.[S39]

Evidence-supported strengths & limits

AdvantagesLimitations / checks
Licensed actor partnerships with published consent and royalty policyCompany policy claims are not the same as an independent talent-rights audit
Built for enterprise narration, shared review and terminology consistencyBusiness/Enterprise capabilities are tier gated and may involve annual commitments
Streaming and async API for product integrationNot a turnkey real-time voice-agent platform
Named L&D and application case studies with concrete workflowsReported speed/ROI outcomes are not independent experiments
Commercial rights listed on paid Studio plansFree trial lacks commercial rights; use is still subject to contract
Documented SOC 2 Type II claims and DPAFull audit findings not examined; private deployment and some technical limits undisclosed
Current Caruso voice directions and pronunciation controlsVoice-dependent expressiveness and pronunciation gaps remain possible

Best suited to: enterprise L&D, instructional design, corporate communications, marketing teams with recurring narrated content, agencies needing quick revisions, and developers embedding authorized speech. Potentially less suitable: unlimited-free TTS seekers, users needing unrestricted identity cloning, teams requiring unsupported local deployment, or projects demanding fully integrated two-way conversational agents without additional systems.[S5] [S7] [S10] [S32]

Frequently asked questions

What is WellSaid?

A privately held U.S. company offering AI voice production tools, licensed synthetic voices, team workspaces and an API for turning text into speech.

Is WellSaid the same as WellSaid Labs?

WellSaid is the current brand. Official legal agreements identify WellSaid Labs, Inc. as the contracting company.

Who founded WellSaid?

Matt Hocking and Mixail/Michael Petrochuk are identified in founding-era investor and press coverage.

When was WellSaid founded?

The current official About page says 2019. Older official LinkedIn information and reporting refer to 2018 research origins; incorporation date is not independently verified.

Where is WellSaid headquartered?

Seattle, Washington, United States, according to the company and LinkedIn.

Who is its current CEO?

Benjamin Dorr, according to WellSaid’s current leadership page.

Is WellSaid public or acquired?

It is publicly described as privately held. No verified parent-company acquisition was established in this review.

What is a Voice Avatar?

WellSaid’s term for a synthetic voice built from approved recordings of a consenting performer; it is not automatically a customer-owned voice clone.

Can WellSaid clone voices?

Its service terms describe consent-based Custom Voice Models that may be made available. Its ethics statement rules out arbitrary, independent cloning without permission.

What is a custom brand voice?

A commissioned or specially arranged modeled voice for an organization, subject to approval, permissions and contractual usage rights.

Does WellSaid have a public API?

Yes. Its TTS API uses authenticated requests and includes synchronous/streaming and asynchronous speech generation workflows.

Does WellSaid support streaming speech?

Yes. Its current API product page explicitly advertises low-latency, real-time streaming TTS.

Is WellSaid a complete conversational AI agent?

No complete agent stack was substantiated. Its API can generate live speech, but a full agent needs dialogue and other layers.

What is Caruso?

Caruso is WellSaid’s documented next-generation voice model, supporting AI Director control via the API, currently for English-language API voices.

How many voices are available?

Current marketing pages show different “240+” and “280+” counts; verify your account’s live voice catalog. Both are vendor-reported.

Which languages does WellSaid support?

The Help Center lists English regional accents plus global options including Arabic, Mandarin, French, German, Japanese, Korean, Spanish, Portuguese and more. Global voices are chiefly Enterprise access.

Can one Voice Avatar switch languages?

The Help Center says voices are trained for specific languages; do not assume the same individual voice supports every language.

How much does WellSaid cost?

At 4 November 2026, the published annually billed Starter and Pro prices were $10 and $33 per month, Business $160 per user per month annually, and Enterprise custom. Month-to-month individual prices are higher.

Is there a free plan?

A free trial tier is listed with 3 downloadable minutes per month and no commercial rights.

Can I use WellSaid audio commercially?

Paid plans advertise commercial rights, subject to the service agreement. The free trial explicitly excludes commercial rights.

Is WellSaid suitable for e-learning?

Yes. Named Microsoft, ARIN, Motul, FlightSafety and other case studies document instructional content use.

Which companies use WellSaid?

WellSaid publishes named stories for Microsoft’s internal enablement team, FlightSafety, ARIN, Motul, 4imprint, TRIPP, PROVOKE and others.

How much has WellSaid raised?

Approximately $12 million in the two reported equity rounds ($2M 2019 seed and $10M 2021 Series A), plus a 2026 debt financing of undisclosed size.

What is WellSaid’s valuation?

A current valuation is not publicly disclosed. Third-party historical figures are not a verified current valuation.

Does it train on customer scripts?

The current Business/Enterprise service agreement says Customer Data will not be used to train WellSaid or third-party AI models. Check the agreement applicable to your account.

Are voice actors compensated?

WellSaid states actors consent in writing and receive recording compensation and usage-based royalties; private contract amounts are not disclosed.

Is WellSaid SOC 2 certified?

WellSaid publicly states SOC 2 Type II compliance and says reports are available on request. This report did not examine the underlying audit.

Does it watermark AI audio?

A universal watermarking or provenance feature was not reliably established from current primary documentation reviewed for this profile.

Is there an official Python SDK?

An official, currently supported Python SDK was not independently confirmed from the reviewed documentation; API docs exist.

How is WellSaid different from ElevenLabs?

WellSaid emphasizes licensed enterprise voiceovers and collaboration; ElevenLabs has broader expressive and agent-oriented speech models. Match the choice to use cases and current terms.

Research notes, disagreements & limitations

TopicWhat this review can responsibly conclude
Research / access date4 November 2026; public pages retrieved close to that date
Current nameWellSaid brand; WellSaid Labs, Inc. legal operator in 2025–26 contracts
Founding chronology2018 research origin vs official 2019 founding; exact incorporation date unresolved
Current leadershipBenjamin Dorr CEO; official leadership page lists eight executives
Latest product update reviewedSeptember 2026 Insights Dashboard export / one-year history; August 2026 voice discovery and Generate All
Model namesCaruso and legacy documented by current API reference; Preview 13.2 still pre-release
Voice cloningConsent-gated contractual custom models; no independent arbitrary-voice cloning
StreamingCurrent public API page affirms streaming and async synthesis
Pricing verified4 November 2026; June 2026 downloaded-minute transition
Latest financing17 March 2026 Multiplier Capital venture debt; amount not published
Valuation / ARRCurrent figures not publicly verified; database guesses not treated as facts
Funding discrepancySome databases list only $10M Series A; 2019 $2M seed is supported by GeekWire
Catalog countsMarketing 240+ vs 280+ voice counts; Help Center says 30+ languages but names fewer top-level language classes
Training datasetActor-consent and private-data claims published; dataset volume/model architecture not disclosed
SecuritySOC 2 Type II vendor statement; audit documentation not independently inspected
SDK / VPC / on-premNot confirmed by the reviewed current public developer/product sources
Customer counts40,000+ vendor-reported customer claim; no independent paid-enterprise count
Internal linksNo Artificial Intelligence Files company/tool/case URLs inserted until they are confirmed to exist
Sources inaccessibleSome detailed trust-center reports and private contract materials require access; investor cap table and full SDK inventory not independently examined

Claims are attributed using evidence categories: Verified primary-document fact (the contract or product page says it); WellSaid-reported (marketing, customer or policy narrative); Customer-reported (quoted named user); Third-party reported (journalism, directory); and Not publicly disclosed / not independently verified where the record is insufficient.[S2] [S3] [S27] [S32]

Sources & references

These are direct links to official agreements, product references, investor and editorial reporting, customer stories and independent reviews. Where a page has no visible author or date, it is marked as undated rather than assigning one. All links were reviewed in the current research pass unless noted otherwise.

Official identity, leadership & legal

S1
About WellSaid: The AI Voice Platform for Enterprises

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/about
S2
Terms of Use

WellSaid Labs, Inc. · 2025-07-09 · accessed 4 Nov 2026

https://www.wellsaid.io/terms-of-use
S3
Services Agreement (Business & Enterprise)

WellSaid Labs, Inc. · 2025-08-13 · accessed 4 Nov 2026

https://www.wellsaid.io/wsa-business-ent
S4
Meet the Leadership Team

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/about/leadership
S12
Data Protection Agreement

WellSaid · 2026-06 update identified · accessed 4 Nov 2026

https://www.wellsaid.io/dpa
S13
Privacy Policy

WellSaid · 2024-07-19 · accessed 4 Nov 2026

https://www.wellsaid.io/privacy-policy
S45
WellSaid Services Agreement (Online)

WellSaid · 2025-08 · accessed 4 Nov 2026

https://www.wellsaid.io/wsa-online
S49
Acceptable Use Policy

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/aup

Products, technical & security

S5
Pricing: Plans Built for Creators, Teams, and Enterprises

WellSaid · Checked 2026-11-04 · accessed 4 Nov 2026

https://www.wellsaid.io/ai-voice-pricing
S6
Understanding June 2026 Pricing and Plan Changes

WellSaid Help Center · 2026-06 · accessed 4 Nov 2026

https://help.wellsaid.io/hc/en-us/articles/52476451642003-Understanding-June-2026-Pricing-and-Plan-Changes
S7
AI Voice & Text-to-Speech API

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/text-to-speech-api
S8
Model Selection with the API

WellSaid API Docs · n.d. · accessed 4 Nov 2026

https://docs.wellsaidlabs.com/reference/model-selection-with-the-api
S9
Release Notes

WellSaid Help Center · through 2026-09 · accessed 4 Nov 2026

https://help.wellsaid.io/hc/en-us/articles/39720425068051-Release-Notes
S10
Ethical & Responsible AI

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/about/ethics
S11
Most Secure AI Voice Generator

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/features/features-security
S14
Voice Actor Job Program

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/voice-actor-jobs
S15
Exploring Languages, Accents, and Styles

WellSaid Help Center · n.d. · accessed 4 Nov 2026

https://help.wellsaid.io/hc/en-us/articles/39079854241299-Exploring-Languages-Accents-and-Styles
S16
List of WellSaid Voices

WellSaid Help Center · n.d. · accessed 4 Nov 2026

https://help.wellsaid.io/hc/en-us/articles/39112914360467-List-of-WellSaid-Voices
S17
AI Voice Generator for Marketing

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/industries/voice-ai-marketing
S40
WellSaid Studio for Adobe Express

WellSaid Help Center · n.d. · accessed 4 Nov 2026

https://help.wellsaid.io/hc/en-us/articles/39187648788243-WellSaid-Studio-for-Adobe-Express
S42
Choosing the Right Studio Plan

WellSaid Help Center · n.d. · accessed 4 Nov 2026

https://help.wellsaid.io/hc/en-us/articles/39149441616531-Choosing-the-Right-WellSaid-Studio-Plan
S43
Home / Current Product Positioning

WellSaid · Checked 2026-11-04 · accessed 4 Nov 2026

https://www.wellsaid.io/
S44
Feature: AI Voices

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/features/ai-voices

Customer / enterprise evidence

S18
WellSaid Case Studies

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/resources/case-studies
S19
FlightSafety International Customer Story

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/resources/case-studies/flightsafety-international
S20
Microsoft Internal Enablement Case Study

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/resources/case-studies/microsoft-internal-enablement
S21
ARIN Streamlines Training Production

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/resources/case-studies/arin-streamlines-training-production-with-wellsaid
S23
S24
PROVOKE Solutions Case Study

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/resources/case-studies/wellsaid-creates-engaging-product-videos-for-provoke
S25
PowerTrain Training Production Case Study

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/resources/case-studies/powertrain-removes-production-bottlenecks-with-wellsaid
S26
4imprint Training Workflow Case Study

WellSaid · n.d. · accessed 4 Nov 2026

https://www.wellsaid.io/resources/case-studies/how-4imprint-built-a-faster-training-with-wellsaid
S41
NPR Planet Money Voice Experiment

WellSaid · 2023-06-06 · accessed 4 Nov 2026

https://www.wellsaid.io/resources/blog/npr-planet-money-host-robert-smith-clones-voice
S51
Global Healthcare Leader: Training Case Study

WellSaid · 2025-09-18 · accessed 4 Nov 2026

https://www.wellsaid.io/resources/blog/global-healthcare-leader-boosts-production-by-50-and-cuts-costs-in-half-with-wellsaid

Funding, investors & external sources

S27
WellSaid Closes Venture Debt Round with Multiplier Capital

WellSaid Labs / PR Newswire · 2026-03-17 · accessed 4 Nov 2026

https://www.prnewswire.com/news-releases/wellsaid-closes-venture-debt-round-with-multiplier-capital-302715429.html
S28
WellSaid Attracts $10M A Round for Higher Quality Synthetic Speech

TechCrunch · Devin Coldewey · 2021-07-07 · accessed 4 Nov 2026

https://techcrunch.com/2021/07/07/wellsaid-attracts-10m-a-round-for-higher-quality-synthetic-speech/
S29
WellSaid Labs Raises $10 Million to Boost Synthetic Voice Business

GeekWire · Alan Boyle · 2021-07-07 · accessed 4 Nov 2026

https://www.geekwire.com/2021/wellsaid-labs-raises-10-million-boost-synthetic-voice-business-fueled-ai/
S30
Portfolio Watch: How WellSaid Labs Is Giving the Future a Voice

Qualcomm Ventures · n.d. · accessed 4 Nov 2026

https://www.qualcommventures.com/insights/blog/portfolio-watch-how-wellsaid-labs-is-giving-the-future-a-voice/
S31
WellSaid Company Profile

LinkedIn · WellSaid · Checked 2026-11-04 · accessed 4 Nov 2026

https://www.linkedin.com/company/wellsaidlabs
S47
WellSaid Labs Company Profile / Historic Funding

Dealroom · Checked 2026-11-04 · accessed 4 Nov 2026

https://dealroom.co/companies/wellsaid-labs/
S48
WellSaid Labs Announces $10M Series A

WellSaid / News Direct · 2021-07-07 · accessed 4 Nov 2026

https://www.nasdaq.com/press-release/wellsaid-labs-announces-%2410m-round-led-by-fuse-to-build-voices-for-every-brand-2021

Reviews & competitive sources

S32
WellSaid Studio Pros and Cons

G2 · Checked 2026-11-04 · accessed 4 Nov 2026

https://www.g2.com/products/wellsaid-studio/reviews?qs=pros-and-cons
S33
WellSaid Studio Reviews

Capterra · 2026-10-19 updated · accessed 4 Nov 2026

https://www.capterra.com/p/238211/Wellsaid/reviews/
S34
Text-to-Speech API and Models

ElevenLabs · Checked 2026-11-04 · accessed 4 Nov 2026

https://elevenlabs.io/docs/overview/models
S35
Resemble Detect

Resemble AI · Checked 2026-11-04 · accessed 4 Nov 2026

https://www.resemble.ai/products/detect
S36
Workspace Guide

Murf · Checked 2026-11-04 · accessed 4 Nov 2026

https://help.murf.ai/what-is-a-workspace
S37
What Is Custom Voice?

Microsoft Learn · Checked 2026-11-04 · accessed 4 Nov 2026

https://learn.microsoft.com/azure/cognitive-services/speech-service/custom-neural-voice
S38
Amazon Polly Generative Engine Update

Amazon Web Services · 2026-10-20 · accessed 4 Nov 2026

https://aws.amazon.com/about-aws/whats-new/2026/10/polly-generative/
S39
Migrate from PlayHT after Shutdown

Inworld AI · 2026-07 · accessed 4 Nov 2026

https://inworld.ai/resources/migrate-from-playht
S50
Murf API Changelog

Murf API · Checked 2026-11-04 · accessed 4 Nov 2026

https://murf.ai/api/docs/changelog