In professional voice production, brand integrity hinges on acoustic consistency. When multi-creator creative teams, corporate learning and development (L&D) departments, and marketing agencies produce hundreds of audio clips across distributed projects, proper nouns, trademarked terms, and industry acronyms face constant phonetic drift. A single brand name pronounced two different ways across consecutive product modules instantly erodes audience trust.
WellSaid Labs resolves this challenge through a centralized, two-pronged phonetic architecture: Pronunciation Libraries (Team Libraries) within WellSaid Studio for interactive creative teams, and Replacement Libraries accessed programmatically via the REST API for scaled production pipelines. Instead of forcing creators to decode complex International Phonetic Alphabet (IPA) strings or construct custom XML Pronunciation Lexicon Specification (PLS) files, WellSaid combines text replacements with a proprietary Respellings system developed alongside Oxford Languages.
Configuring this shared pronunciation environment correctly allows creative and production teams to automate terminology enforcement, eliminate post-production rework, and maintain an identical brand voice across all collaborative workspaces.
Core Mechanics: Replacements, Respellings, and Shared Libraries
Maintaining brand consistency begins with understanding how WellSaid handles phonetic substitutions under the hood. The platform distinguishes between basic text swaps and phonetic coaching, structuring them into personal and team-wide repositories.
+-------------------------------------------------------------------------------+
| WELLSAID LABS STUDIO |
| |
| +------------------------+ +--------------------------+ |
| | Personal Library | | Team Library | |
| | (Creator-specific | | (Shared Team Workspace | |
| | exploratory rules) | | Universal Brand Rules) | |
| +-----------+------------+ +-------------+------------+ |
| | | |
| +-------------------+ +------------------------+ |
| | | |
| v v |
| +-----------------------------+ |
| | Active Studio Script | |
| | (Oxford & Smart Suggestions | |
| | Inline ::RESPELL::) | |
| +--------------+--------------+ |
+------------------------------------|------------------------------------------+
|
v
+------------------------------------+------------------------------------------+
| AUTOMATED PRODUCTION & REST API |
| |
| POST /v1/tts/stream |
| { "text": "...", "speaker_id": 7, "library_ids": "brand-library-uuid" } |
| |
| +-------------------------------------------------------------------------+ |
| | API Replacement Libraries Repository | |
| | - "original": "Kubernetes", "replacement_text": "koo-ber-NET-eez" | |
| | - "is_phonetic_respelling": true, "enabled": true | |
| +-------------------------------------------------------------------------+ |
+-------------------------------------------------------------------------------+
Text Replacements vs. Phonetic Respellings
WellSaid provides two complementary methods for altering how text is spoken:
- Text Replacements (
is_phonetic_respelling: false): Instructs the text-to-speech (TTS) engine to substitute the original text with alternative written words before synthesis. This is optimal for acronyms meant to be pronounced letter-by-letter (such as expanding "AWS" to "A W S") or translating non-standard symbols and technical unit codes without altering the written script displayed to editors. - Phonetic Respellings (
is_phonetic_respelling: true): Employs WellSaid’s custom grapheme-to-phoneme system to dictate exact syllable articulation and lexical stress. Rather than relying on 107 IPA symbols and cumbersome diacritics, WellSaid simplifies English phonetics into 45 core sound pairings (12 vowels, 6 vowel combinations, and 27 consonants). Syllables are separated by hyphens, and capitalized syllables signal primary stress (for example,PEET-suhfor "pizza" orsask-ACH-oo-whenfor "Saskatchewan").
Personal Library vs. Team Library
In WellSaid Studio, every user account includes a Personal Library. Rules saved to a Personal Library apply only to that specific creator's individual projects and takes.
On Team, Business, and Enterprise plans, organizations receive a shared Team Library within their centralized Team Workspace. When a rule is committed to the Team Library, WellSaid Studio automatically evaluates and applies that replacement across every script created within the shared workspace. Any creator opening a project or rendering a voiceover clip instantly inherits the standardized pronunciation without manually importing rules or memorizing pronunciation guides.
Step-by-Step Configuration: Setting Up a Shared Team Library
Deploying a shared pronunciation library across collaborative production teams requires a structured workflow: discovering ambiguous terms, crafting and auditing phonetic respellings, publishing to the shared workspace, and enforcing governance.
Step 1: Terminology Audit and Script Highlighting
Before generating audio, creative leads should compile an official brand terminology list. This includes proprietary product names, partner brands, technical acronyms, and executive surnames.
Within the WellSaid Studio editor:
- Double-click or highlight the target word in the text block.
- The contextual Smart Toolbar appears above the selection.
- Review the automated recommendations:
- Oxford Suggestions: Pre-calculated phonetic options sourced from WellSaid’s partnership with Oxford Languages, mapping the syllabified lexical database directly into respelling syntax.
- Smart Suggestions: AI-generated phonetic predictions tailored to conversational context.
- If an Oxford or Smart Suggestion matches the desired brand pronunciation, select it. If the term is proprietary or requires unique emphasis, click + New Replacement.
Step 2: Crafting and Auditioning Custom Respellings
When creating a custom replacement, team members must adhere to WellSaid’s phonetic syntax to prevent synthesis artifacts:
- Syllable Separation: Every syllable must be divided with a single hyphen (
-). - Lexical Stress: Capitalize all letters of the syllable receiving primary stress (e.g.,
al-GOR-ih-dhuhmvs.AL-guh-rih-dhuhm). Unstressed syllables must remain strictly in lowercase. - Allowed Characters: Use only Latin alphabetic letters and hyphens. Do not introduce numbers, punctuation marks, or mixed-case lettering within a single syllable.
- Phonetic Reference Mapping: Match complex vowels to WellSaid’s standard guide (e.g.,
AYfor long a as in eight,EEfor long e as in ease,OHfor long o as in oat,OOfor ooh, andUHfor short u as in up).
+-------------------+--------------------+------------------------+-------------------------------------+
| Term Category | Original Text | WellSaid Respelling | Phonetic Rationale |
+-------------------+--------------------+------------------------+-------------------------------------+
| Tech Brand | Asana | uh-SAH-nuh | Emphasizes middle open syllable |
| SaaS Terminology | SaaS | SASS | Spoken as a unified word, not letters|
| Enterprise Cloud | Kubernetes | koo-ber-NET-eez | Third-syllable primary stress |
| Hardware Protocol | PCIe | P-C-I-E | Enforces distinct initialisms |
| Complex Geography | Saskatchewan | sask-ACH-oo-when | Prevents slurred mid-vowel sounds |
+-------------------+--------------------+------------------------+-------------------------------------+
Production Tip: Always audition the phonetic respelling across multiple voice avatars prior to saving. Because voice models possess distinct baseline cadence, pitch contour, and accent characteristics (such as US General American versus British accents), previewing ensures the respelling sounds natural across diverse voice talents.
Step 3: Saving to the Shared Team Library
Once the auditioned replacement produces the exact brand vocalization:
- In the Smart Toolbar replacement panel, ensure the Phonetic Respelling toggle is switched ON (leave it OFF only if performing a direct literal text swap).
- Choose Save to Team Library (rather than Personal Library).
- The replacement is immediately activated across all team members who operate within that workspace. Future occurrences of the original word in any project will display the replacement badge and render using the verified pronunciation.
Workspace Governance and Team Collaboration Settings
Maintaining pronunciation consistency across large teams requires access control and administrative discipline to prevent conflicting entries.
Project Sharing and Role Permissions
WellSaid Studio provides tiered permissions across projects and workspaces:
- Team Workspaces: On Business and Enterprise tiers, all folders and projects created within the shared Team Workspace are accessible by all team members by default, ensuring immediate inheritance of shared dictionaries.
- Project-Level Permissions: For sensitive or unreleased campaigns, projects can be initialized in personal workspaces and shared with granular controls:
- Full Access: Allows team members to edit scripts, manage takes, configure library rules, and re-share the project.
- Can Edit: Permits team members to draft copy, select voice avatars, and render takes, but restricts administrative folder management.
- Can View: Restricts collaborators or external clients to auditioning takes and reviewing generated audio without modifying scripts or library settings.
Designating a Pronunciation Steward
In collaborative production environments, multiple creators independently editing rules can lead to conflicting phonetic overrides (e.g., one creator respelling a brand name as ZEE-roh and another as ZEER-oh). Organizations should designate an editorial lead or voice producer as the Pronunciation Steward.
The Steward oversees:
- Bi-weekly reviews of the Team Library to eliminate redundant or conflicting entries.
- Approving global overrides for cornerstone brand terminology.
- Distinguishing between global rules and project-specific variations.
Managing Terminology at Scale: CSV/JSON Bulk Workflows and the REST API
High-volume production environments—such as e-learning localization pipelines or dynamic marketing video rendering—cannot rely solely on manual GUI entry. WellSaid Labs provides programmatic and bulk-file capabilities to synchronize dictionaries across enterprise ecosystems.
Bulk Import and Export via CSV
WellSaid Studio supports importing and exporting entire pronunciation libraries using standard CSV files:
- Exporting for Audits: Creative teams can export the current Team Library to inspect, clean, and document phonetic assets in spreadsheet software.
- Enterprise Glossaries: Corporate terminology bases (from tools like SDL MultiTerm, Phrase, or internal glossaries) can be formatted into WellSaid-compliant CSVs (mapping
Original Word,Replacement, andIs Phonetic Respelling) and imported directly into the Team Workspace. - Version Control: Storing exported CSV dictionaries in a version-controlled repository (such as Git) provides an immutable audit log of how brand pronunciation guidelines evolved over time.
Automated Pipelines: REST API Replacement Libraries
For engineering and media production teams integrating WellSaid into automated rendering stacks, WellSaid Labs exposes dedicated /v1/tts/replacement_libraries endpoints.
1. Creating a Dedicated Replacement Library
Production teams can isolate dictionaries per client, campaign, or product line:
curl --location 'https://api.wellsaidlabs.com/v1/tts/replacement_libraries' \
--header 'X-API-KEY: YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"name": "Global Brand Terminology 2026"
}'
Response returns a unique library_id.
2. Programmatically Injecting Terminology
Populate the library with approved phonetic respellings:
curl --request POST \
--url https://api.wellsaidlabs.com/v1/tts/replacement_libraries/YOUR_LIBRARY_ID/replacements \
--header 'X-API-KEY: YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"original": "WellSaid",
"replacement_text": "WEHL-sehd",
"is_phonetic_respelling": true,
"enabled": true
}'
3. Automated Respelling Lookup via API
When ingesting third-party copy, backend pipelines can query WellSaid’s lexical database to fetch Oxford-backed recommendations before pushing text to synthesis:
curl --request GET \
--url 'https://api.wellsaidlabs.com/v1/tts/respelling_suggestions?word=charcuterie' \
--header 'X-API-KEY: YOUR_API_KEY'
Returns candidate phonetic respellings such as SHAR-KOO-dur-ee.
4. Applying Libraries During Synthesis
To enforce the shared dictionary during automated voiceover generation, include the library_ids parameter in the TTS rendering call:
curl --request POST \
--url https://api.wellsaidlabs.com/v1/tts/stream \
--header 'X-API-KEY: YOUR_API_KEY' \
--header 'Content-Type: application/json' \
--data '{
"speaker_id": 7,
"text": "Welcome to our product enablement course on WellSaid Studio.",
"library_ids": "YOUR_LIBRARY_ID"
}' > voiceover.mp3
WellSaid automatically evaluates the input text against the referenced replacement library, performs the phonetic substitutions, and passes the formatted tokens to the synthesis model.
Critical Traps and Operational Best Practices
Configuring shared pronunciation dictionaries introduces several subtle pitfalls that can compromise audio quality if left unmanaged.
1. Beware the Homograph Trap
The most dangerous mistake in shared dictionary configuration is assigning global rules to homographs—words that share identical spelling but carry different pronunciations and meanings depending on context.
Common examples include:
- "Live": Verb ("They live in Boston" -> short i) versus Adjective ("Broadcast live from Seattle" -> long i,
LYV). - "Read": Present tense ("I read daily" ->
REED) versus Past tense ("She read the report" ->REHD). - "Record": Noun ("Set a new record" ->
REH-kerd) versus Verb ("Record the clip" ->rih-KORD).
If a team member creates a global Team Library rule replacing "live" with LYV, every narration script using the verb "live" will sound completely broken.
Best Practice: Never place homographs in a global Team Library. Instead, handle context-dependent words through local, inline tags directly in the script editor:
- Standard Model Syntax: Wrap the respelling in double colons:
Coming to you ::LYV:: from headquarters. - BETA Model / AI Director Syntax: Use semantic markup:
Coming to you <respell value="LYV">live</respell> from headquarters.
2. Disabling vs. Deleting Entries
When a product line undergoes rebranding or a temporary promotional pronunciation expires, do not immediately delete the entry from the library. In the REST API, toggle "enabled": false. In Studio, deactivate the replacement rule. This preserves the verified phonetic engineering for future archival rendering while preventing it from altering new scripts.
3. Dialect and Accent Sensitivity
While WellSaid’s Respellings system is primarily grounded in US General American phonology, global creative teams frequently deploy British or international voice avatars. Certain respellings optimized for rhotic American accents (such as syllables ending in -er) may render unnaturally with non-rhotic British voice models. Production teams should audit key brand terms against their designated regional avatar rosters before standardizing entries across global sub-teams.
Summary Checklist for Collaborative Production
To maintain brand terminology consistency across all projects, creative and production teams should implement this operational cadence:
- Centralize Workspaces: Ensure all active project creators belong to the official WellSaid Team Workspace under a Business or Enterprise subscription.
- Standardize Core Jargon: Populate the Team Library with foundational brand names, acronyms, and executive leadership names using Oxford Suggestions and auditioned phonetic respellings.
- Isolate Homographs: Ban homographs from team-wide libraries; mandate inline
::RESPELL::or<respell>syntax for dual-pronunciation words. - Synchronize Production Stacks: If operating automated rendering pipelines, link API requests directly to the corresponding
library_idsto ensure parity between human editors and batch rendering engines. - Periodic Auditing: Have the designated Pronunciation Steward export library CSVs monthly to prune deprecated terms, eliminate duplicates, and resolve phonetic inconsistencies.