f4x Automatic
Speech Recognition
„…best detection performance in the test …“
c´t 17/21
„…best detection performance in the test …“
c´t 17/21
For interviews, podcasts, zoom recordings… etc. f4x runs directly in the browser. Your uploaded data is secure and GDPR compliant. Every new registered account gets 60 minutes of speech recognition for free.
For recording in 48 languages…
c’t already certified 2021 “best detection performance in the test”.
(c´t 17/21). In October 2022, we completely renewed f4x and improved the recognition by about 50%. Good transcripts are now generated even from challenging shooting situations. This significantly reduces correction times.
Test f4x now with your data.
We give you 60 minutes contingent for new registration.
We pay attention to high security standards and transparent infrastructure so that your data is processed by us in compliance with GDPR.
Your data will be used exclusively for speech recognition and not for other purposes or even passed on. The transmission of your data is already encrypted during the upload.
Read our f4x procedure concept (submitted to the Hessian Data Protection Commissioner) here.
Speech recognition runs exclusively on ISO-27001 certified servers in Germany.
Immediately after the conversion, the recordings are deleted. The text file is stored in encrypted form until collection. After that, the data is deleted from our servers. The finished transcripts are then only available locally to you as a text file.
By the way, our servers not only run securely, but 100% with climate-neutral energy from hydropower.
No subscriptions,
No time limit,
Billing by the minute according to material length,
contingent can be used in any time.
from 5,86 € per hour
Top
Package 100 hours |
||||||
---|---|---|---|---|---|---|
2 hours | 5 hours | 15 hours | 50 hours | 200 hours | ||
studying / doctorate | 22 € 11 €/h |
44 € 8,80 €/h |
88 € 5,86 €/h |
|||
all others | 33 € 16,50 €/h |
55 € 11 €/h |
144 € 9,60 €/h |
444 € 8,88 €/h |
888 € 8,88 €/h |
1555 € 7,78 €/h |
Prices incl. VAT
Voice recognition requires a quota, which you can buy in hourly packages and redeem in your account.
Billing is per minute per uploaded material,
the quota can be used flexibly and does not expire.
No subscriptions or similar standing payments.
If you purchase 100 hours or more, you can receive them on request as any combination of individual contingent codes. Flexible and can be used independently of time. You can use these codes in your project or pass them on to external employees or students.
We are also happy to set up domain accounts for institutions. All users with a common e-mail domain, e.g. …@stud.uni-example.com can use a common quota. Uncomplicated to set up, no further administration necessary.
Please contact us by email before purchase.
Manchmal kann es mehrere Stunden dauern, bis eine Datei vollständig hochgeladen ist. Das hat nicht unbedingt etwas mit der Länge der Datei zu tun, sondern meist mit der Dateigröße. Wenn Ihre Dateien größer als 1 GB (1000 MB) sind, ist es ratsam, sie vor dem Hochladen zu verkleinern. Dies kann z.B. durch das Umwandeln einer Videodatei in eine reine Audiodatei oder durch das Komprimieren einer großen Audiodatei in mp3 mit 192kBit geschehen (bspw. mit Switch von NCH). Sollte der Upload dennoch mehrere Stunden dauern, z.B. weil Ihre Internetverbindung sehr langsam oder unser Server überlastet ist, kann es sein, dass Sie ungewollt ausgeloggt werden und der Upload nicht erfolgreich abgeschlossen wurde. Starten Sie in diesem Fall den Upload zu einem späteren Zeitpunkt erneut und versuchen Sie ggf. eine schnellere Internetverbindung oder eine kleinere Dateigröße.
Unsuccessful uploads are not charged, but are initially noted as reserved. The credit will be credited back to you after a maximum of 24 hours. You can then start the upload again at a later time.
If you have not yet downloaded your transcripts and need to change your password, these transcripts will be deleted. As your files are encrypted with your old password, existing transcripts can no longer be opened after resetting the password. Consequently, all existing transcripts will be deleted when the password is reset, as no one will be able to read them. Any lost quota will be booked back to you and you can have the transcripts generated again at no extra cost. We take the security of transcript data very seriously. In this case, this means a small loss of convenience, but significantly more security for your data.
Some audio or video files have been created in a format that is not supported by f4x. Unfortunately, this is possible regardless of the file extension. For example, an Mp3 file may have been created in very different ways. If your files are not supported by f4x, you can convert them using the freeware version of NCH Switch or numerous other programmes. Stereo MP3 with 192kBit fixed (not variable) bit rate has proven to be quite reliable here.
Very rarely (< 1%) does our speech recognition system fail to recognise that different speakers can be heard on a recording (e.g. in interviews with twins). In such a case, the transcript looks like a very long continuous text without paragraph breaks. Even uploading the recording again would not produce a better result. In this case, it is necessary to insert paragraphs automatically during the correction process wherever there has been a change of speaker and to insert the names of the speakers and, if necessary, time stamps. The easiest way to do this is directly in f4transcript – the names of the speakers can be entered there and automatically inserted alternately when there is a paragraph change. Timestamps are then also automatically entered at the end of the paragraph.
When uploading a file to f4x, you must specify which language is to be recognised in this recording. If the language changes in a recording, e.g. constantly from German to English and back to German, f4x cannot implement this correctly and will only recognise the language you have preselected and potentially leave other parts blank or insert unsuitable content.