About
AI tools for people who publish audio and video
SpeakScribe turns recordings into clean text, text into natural-sounding speech, and finished videos into publish-ready YouTube metadata — in one place, from $0.
What SpeakScribe does
Transcription
Upload a recording and get an editable transcript with timestamps. Speech recognition supports 99 languages.
Voice Studio
Turn text into natural-sounding speech with 361 voices in 75 languages, including regional accents.
YouTube SEO
Generate titles, descriptions and tags from your transcript so a video is ready to publish.
Our guides explain how transcription behaves for specific languages, file types and real situations, including where automatic speech recognition is likely to make mistakes.
Who runs SpeakScribe
SpeakScribe is built and run by the SpeakScribe Team. We develop the product, write the guides and articles on this site, and answer support email ourselves at audioflowteam@speakscribe.me.
Who it is for
Podcasters, YouTubers, journalists, students, researchers and small teams who need accurate text from spoken audio without paying for several separate tools. You can try transcription and voices without creating an account.
How we treat your files
- Your uploaded audio and text are used only to produce the result you asked for.
- We do not use your content to train AI models and we never share it with advertisers.
- Files are deleted automatically after the retention period of your plan (7 days on the free plan).
- Read the full details in our Privacy Policy and Terms of Service.
Honest about accuracy
Automatic transcription is very good on clear speech and less reliable on heavy background noise, overlapping speakers and rare names. We tell you where each language struggles, and every transcript can be downloaded as plain text so you can correct it before you export.
What runs under the hood
- Transcription uses Whisper large-v3-turbo, OpenAI's open speech recognition model. If that service is busy, a backup engine finishes the job. How transcription works.
- Voices come from two sources: Studio voices run on our own servers with the open Kokoro model (adjustable speed, WAV), and Global voices are Microsoft neural voices (MP3). Together they cover 75 languages.
- Publish packages, spelling fixes and translation are written by a large language model from your own transcript or script.
How we write our guides
- Guides on this site — language pages, file-type pages and articles — are written for one purpose: helping you get a better transcript or voiceover in that specific situation.
- Product numbers (limits, prices, voice counts) are taken from the product itself and updated when it changes. Platform rules, such as YouTube's metadata limits and policies, link to the platform's own help pages.
- We don't publish accuracy percentages, invented statistics or reviews. Where we haven't measured something, we say what affects it instead.
- Spotted a mistake, or a tip that is wrong for your language? Tell us through the contact page and we will correct it.
Pricing
There is a free plan with monthly allowances for transcription, voices and publish packages. Paid plans start at $4.99 per month and are billed through PayPal, or by card, Google Pay, Apple Pay or UPI through Dodo Payments; you can cancel at any time. See Pricing for the current limits.
Get in touch
Questions, bug reports, billing help or data requests — see our Contact page.