Custom Dictionary for Dictation: Teach SuperSpeech Your Vocabulary
Map spoken variants to preferred output. Brand names, technical terms, legal Latin, medical jargon — SuperSpeech learns your vocabulary.
Every profession has vocabulary that trips up generic speech recognition. Developers say "kubectl," doctors say "fluoxetine," lawyers say "res ipsa loquitur." The SuperSpeech custom dictionary solves this by letting you map any number of spoken variants to the desired output text — accurately, case-sensitively, and locally.
How it works
You define dictionary entries in a simple JSON structure. Each entry has an output (what should appear) and a list of variants (how people actually say it). When any variant is detected in transcription, it is replaced with the output automatically. Case sensitivity is toggleable per entry.
- Unlimited entries — no subscription tier limits
- Case-sensitive or case-insensitive per entry
- Per-language variants supported
- Import/export JSON for sharing across a team
- Enable/disable entries without deleting
Common dictionary use cases
Use cases we see most often:
- Brand names: "super speech" → "SuperSpeech"
- Acronyms: "a p i" → "API", "s q l" → "SQL"
- CLI tools: "kube control" → "kubectl", "git hub" → "GitHub"
- Legal Latin: "ex parte" → "ex parte" (preserves italics in DOCX)
- Drug names: "zolpidem" → "Zolpidem"
- Client and case names
JSON format example
A dictionary is a JSON array. Example:
File locations
On Mac the dictionary lives at ~/Library/Application Support/SuperSpeech/custom_dictionary.json. On Windows at %LOCALAPPDATA%/SuperSpeech/custom_dictionary.json. You can also pass a dictionary to the TranscribeCLI via the --dictionary flag.
Frequently Asked Questions
Is there a limit on entries?
No. Unlimited dictionary entries on all plans. Large enterprises have loaded tens of thousands of entries without performance issues.
Can I share a dictionary with my team?
Yes. Export the JSON file and anyone on your team can import it. Git-committable for version control.
Does it work per language?
Yes. You can specify which language an entry applies to, or leave it language-agnostic. Useful for multilingual professionals with language-specific terminology.
Can I use regex in variants?
No. Variants are plain strings for simplicity and predictability. For advanced normalization, you can chain dictionaries or post-process with a grammar LLM.
What if a variant overlaps with regular words?
The matcher prefers exact token matches and respects word boundaries. If "api" would collide with a real word in your language, set case-sensitive to true or add context to the variant (e.g., "the a p i" instead of "a p i").
Teach SuperSpeech your vocabulary
Custom dictionary is included in every plan. Start mapping terms in 5 minutes.