Convert SPH to FLAC
Turn SPH recordings into FLAC. Compressed without loss as FLAC, at 16 or 24 bits as the source or the edits need unless audio.bit_depth is given.
Converted in an isolated sandbox on our serversDeleted after 24 hours
Drop SPH files on the dialor turn it to pick what to do
Speech recordings as lossless FLAC.Your SPH file becomes one FLAC file that keeps the sound exactly.
What you get
One FLAC file made from the NIST SPHERE recording. FLAC is lossless compressed audio, so the sound is stored without losing detail, and it takes about half the size of WAV.
The input's bit depth is kept unless you set Bit depth to 16 or 24.
Shaping the sound
Several options change the audio on the way. Set Channels to 1 to mix the sound down to mono, which suits speech. Set Reduce noise to lower steady background noise such as hum, hiss and fans. Sudden noises such as a door are left in.
Set Normalize loudness to bring quiet and loud recordings to -16 LUFS. Normalizing also resamples to 44.1 kHz unless you set Sample rate (Hz). Set Pitch to move the pitch by whole semitones at the same speed, or set Speed to play faster or slower while voices keep their pitch.
Cutting the recording
Set Start at (s) to cut seconds from the beginning. Decimals such as 1.5 work. Set Stop at (s) to end at a second of the original. It counts from the start of the original, not from where trimming starts, so it must be later than Start at. To cut a FLAC file after converting, see trim FLAC.
SPH and FLAC.What each format is, and the options this conversion takes.
- 01 · audio.bit_depth
Bit depth
The precision of each sample in uncompressed and lossless audio. 16 is CD quality, and 24 leaves room for editing. The input's depth is kept when it is left out.
16 · 24 - 02 · audio.channels
Channels
1 mixes the sound down to mono, which halves its size and suits speech. 2 keeps it stereo.
1 to 8 - 03 · audio.denoise
Reduce noise
Lowers steady background noise such as hum, hiss and fans. Sudden noises such as a door are left in.
- 04 · audio.normalize
Normalize loudness
Makes quiet and loud recordings play at the same loudness, -16 LUFS, the level podcasts and streaming services aim for.
- 05 · audio.pitch
Pitch
Moves the pitch up or down by whole semitones while the speed stays the same. 12 is one octave up and -12 one octave down.
-12 to 12 - 06 · audio.sample_rate
Sample rate (Hz)
How many times per second the sound is measured. 44100 is the CD rate and 48000 the video rate. Lower makes a smaller file and dulls high sounds.
8000 to 192000 - 07 · speed
Speed
Plays faster or slower. 2 is twice as fast and 0.5 half as fast. Voices keep their pitch.
0.25 to 4 - 08 · trim_end
Stop at (s)
Stops at this second of the original. It counts from the start of the original, not from where trimming starts, so it must be later than Start at.
0 to 86400 - 09 · trim_start
Start at (s)
Cuts this many seconds from the beginning. Decimals such as 1.5 work.
0 to 86400
The formats
NIST SPHERE Audio
Speech recordings in the NIST SPHERE format of the TIMIT and LDC speech corpora.
- Extensions
- .sph
- MIME type
- audio
/x -nist -sphere - Kind
- Audio
FLAC Audio
Lossless compressed audio. Exact copies of CDs at about half the size of WAV.
- Extensions
- .flac
- MIME type
- audio
/flac - Kind
- Audio
In your code.The same SPH to FLAC conversion through the API, for your app or your agent.
$ cv convert recording.sph recording.flac$ cv convert recording.sph recording.flacQuestions.What people ask about SPH to FLAC.
FLAC is lossless, so the sound is stored exactly. Options such as Reduce noise, Normalize loudness, Pitch and Speed change the sound only when you use them.
Normalize loudness also resamples to 44.1 kHz. Set Sample rate (Hz) to choose another rate.
Yes. Set Channels to 1 to mix the sound down to mono. This halves its size and suits speech.
Yes. Set Start at (s) and Stop at (s) to keep only the part you want.