Description
The problem
I need to generate MP3 with bitrate exactly 48kbps. For this purpose, I'm trying to use TTS API method "/api/tts_get_url" with additional options
NOTE WHY: The 48kbps (exactly!) is THE ONLY format that can be played on Alexa Media Player by putting url in "<audio src=..."> tag. This is a workaround forcing Alexa to play announcements in non-Alexa-friendly language.
I have noticed that "/api/tts_get_url" request can be extended with options forcing MP3 conversions:
"preferred_format"
"preferred_sample_rate"
"preferred_sample_channels"
"preferred_sample_bytes"
BUG:
I'm sending TTS API request with data containing
{
'engine_id': 'microsoft',
'message': 'Any message',
'options': {'preferred_sample_rate': '16000', 'preferred_sample_channels': 2, 'preferred_sample_bytes': 2}
}
the TTS code converts preferred_sample_bytes option to ffmpeg parameter using this code
if to_sample_bytes == 2:
# 16-bit samples
command.extend(["-sample_fmt", "s16"])
and then the exception is raised:
Stream mapping:
Stream #0:0 -> #0:0 (mp3 (mp3float) -> mp3 (libmp3lame))
[libmp3lame @ 0x7f96dc9d1000] Specified sample format s16 is not supported by the libmp3lame encoder
[libmp3lame @ 0x7f96dc9d1000] Supported sample formats:
[libmp3lame @ 0x7f96dc9d1000] s32p
[libmp3lame @ 0x7f96dc9d1000] fltp
[libmp3lame @ 0x7f96dc9d1000] s16p
EXTENSION (PLEASE):
Using current set of options I'm still unable to convert MP3 to bitrate exactly 48kbps. It looks like ffmpeg requires exact audio bitrate parameter -b:a
So... Could you add addional option of TTS API
preferred_bitrate
so it could send request like
{
'engine_id': 'microsoft',
'message': 'Any message',
'options': {'preferred_sample_rate': '16000', 'preferred_sample_channels': 2, 'preferred_bitrate': '48k'}
}
to enforce ffmpeg encoder to use desired bitrate on output
if to_bitrate is not None:
command.extend(["-b:a", str(to_bitrate)])
The ffmpeg command I'm expecting to get (tested that produces MP3 acceptable by Alexa Media Player) is something like:
ffmpeg -y -i {input} -ac 2 -codec:a libmp3lame -b:a 48k -ar 16000 -f mp3 {output.mp3}
What version of Home Assistant Core has the issue?
core-2025.1.4
What was the last working version of Home Assistant Core?
No response
What type of installation are you running?
Home Assistant OS
Integration causing the issue
TTS
Link to integration documentation on our website
https://www.home-assistant.io/integrations/tts/
Diagnostics information
Example YAML snippet
Anything in the logs that might be useful for us?
Stream mapping:
Stream #0:0 -> #0:0 (mp3 (mp3float) -> mp3 (libmp3lame))
[libmp3lame @ 0x7f96dc9d1000] Specified sample format s16 is not supported by the libmp3lame encoder
[libmp3lame @ 0x7f96dc9d1000] Supported sample formats:
[libmp3lame @ 0x7f96dc9d1000] s32p
[libmp3lame @ 0x7f96dc9d1000] fltp
[libmp3lame @ 0x7f96dc9d1000] s16pAdditional information
No response