We made free telegram botwhich converts audio to text. Unlike our free public solutions for transcribing long audio, this bot is rather configured for the convenience of working with short voice messages, notes and audio of medium length (several minutes).
You can send audio to the bot either directly or add it to a group. In the group, the bot will respond to all audio files (but no error messages will be displayed). You can learn more about the limitations and peculiarities of work in the methods
The main UX feature of the bot’s work is a well-developed and convenient format for reading and navigating recognized messages and notes (and not a wall of text).
How to use
You just need to walk along link or just open the bot page
@silero_audio_bot in the Telegram. Then:
- Enter the command
/faqhelp you understand the basic assumptions and limitations in the work;
We have thoroughly tested the main limitations and compiled the list of main problems and edge cases into the commands described above. If you find new edge cases, direct contacts are indicated there, you can write to us. We do not hide or bite.
Key features and differences from existing solutions
Immediately so that you understand – we develop algorithms, and not just transfer your personal data to “jaysons” in the API of corporations. All algorithms and developments involved in the work of the bot are our own.
We are proud that:
- The bot works on the basis of our systems speech detection (by the way, a huge update is coming soon!), speech recognition, punctuation and capitalization;
- Our recognition system is at least not inferior to the solutions of corporations (a debatable question, according to our research on about 20 different domains, our system was the best on most of them, except for a number of striking exceptions, such as calls to banks);
- The bot “tries” to be as user-friendly as possible in terms of UX and readability within the Telegram capabilities;
We DO NOT do the following things:
- We DO NOT collect user data;
- We are NOT in the business of “selling audiences” to investors or third parties;
- We DO NOT send your data to the IPA of corporations or companies affiliated with oligarchic structures;
- We DO NOT appropriate others’ achievements, all our own achievements;
Security and restrictions
We did not remove the limitation of 20 megabytes per file for Telegram, because in a normal audio format, 20 megabytes can fit long enough audio for a given UX scenario, and I repeat that there is a separate one. service for “long” audio to be developed separately. From the point of view of user convenience, it is strange to solve desktop use cases in the messenger.
Each user has a limit, we block bots and prevent inappropriate use using the methods available to us.
We completely encrypt all traffic and data. We do not “donate” your personal information to corporations or oligarchic (or their affiliated) structures.
As well as our solution for transcribing long files, we will develop and improve our solution if it is useful to the people. In a sense, the fate of the bot and the availability of an independent tool are in your hands.
We tried to do everything beautifully, but if there are any bugs and problems, we will be grateful to the community for help. At first, there may be slight interruptions in the work when rolling out fixes, but we hope that we have already debugged 95% of all edge cases, and the remaining 5% can be quickly corrected or understood and forgiven.