Is Bulgarian supported by your engine?
Not by ours. We accept 99+ languages, and the service contract is explicit. Bulgarian audio belongs to a service that lists Bulgarian.
Transcribe Bulgarian audio to text: our engine runs 99+ languages. The word-error class AssemblyAI publishes for Bulgarian, the services that list it and what they charge, read on 21 and 22 October 2026.
Bulgarian is not one of our 99+ languages: we run 99+ languages only. AssemblyAI rates Bulgarian in the class of 10–25 % word error, Sonix lists it among 54 transcription languages, and Descript only translates captions into Bulgarian. Our file limit is 30 minutes.
Bulgarian is one of the languages the market handles with numbers rather than adjectives, and the numbers are good: AssemblyAI puts Bulgarian in the class below 25 % word error, next to Czech, Danish and Romanian. We keep our own list short, 99+ languages and 99+ languages, and we publish the class tables we read instead of a marketing sentence.
| What | Value |
|---|---|
| Our engine | 99+ languages; Bulgarian is not one of them |
| AssemblyAI, Universal-2 | 99+ languages; Bulgarian sits in the class of 10–25 % word error |
| Sonix | 54 transcription languages, Bulgarian among them; a Bulgarian page in its sitemap |
| Descript | 61 caption-translation and 30 dubbing languages include Bulgarian; its 25 transcription languages do not |
| HappyScribe | 150+ languages and dialects, and a transcribe-bulgarian page in its sitemap |
| TurboScribe | 98+ transcription languages, and a transcribe-bulgarian-audio-to-text page in its sitemap |
AssemblyAI's model reference lists Bulgarian among the languages with a 10–25 % word error rate, next to Arabic, Czech and Romanian. That is a measured number, read on 21 October 2026.
Sonix publishes 54 transcription languages and Bulgarian is on it, together with a Bulgarian page in its sitemap. One list and one page for the same language.
Bulgarian is not among the 25 languages Descript transcribes. It is one of 61 for caption translation and one of 30 for dubbing, which is a different job.
A word-error class is the vendor's own measure of its own engine. It says nothing about your recording: a noisy room in Sofia is still a noisy room.
Our part of the job: up to 30 minutes, $0 for 3 files a day, 6 export formats. Everything comes back with timecodes and, on request, the speakers named.
A measured error class beats a language count.
The class describes the engine, not your room.
Ask which unit the price counts.
Try the worst minute of your own recording.
Not by ours. We accept 99+ languages, and the service contract is explicit. Bulgarian audio belongs to a service that lists Bulgarian.
Its own table puts Bulgarian in the 10–25 % word error class, with Arabic, Czech, Danish and Romanian. Read on 21 October 2026.
Not in its 25 transcription languages. It does translate captions into Bulgarian, one of 61 languages, and dubbing covers 30 languages including Bulgarian.
Sonix bills $10 an hour, or Core at $25 a month. TurboScribe is $10 a month paid yearly and Notta $8.17 a month paid yearly. Three different units.
Not in one file for us. We take 99+ languages, up to 30 minutes on the free plan. Send the Bulgarian half to a service that names the language.
Which languages our engine takes, with the code of each and whether the language is detected automatically.
Transcribe Serbian audio to textA language from the same shortlist.
Transcribe Lithuanian audio to textThe middle of the error tables.
Transcribe English audio to textThe 99+ languages our own engine runs.
Speakers by name, a time on every line, export to TXT, SRT, VTT, CSV and Markdown.