ElevenLabs Opus 48 kHz output vs Sume TTS mp3, wav and raw formats

What ElevenLabs lists

The ElevenLabs text to speech page lists these output families: MP3 from 22.05 to 44.1 kHz at 32 to 192 kbps, PCM S16LE from 16 to 44.1 kHz, mu-law and A-law at 8 kHz, and Opus at 48 kHz with 32 to 192 kbps.

If your product plays audio through an Opus-native stack, getting Opus directly saves a conversion step.

What Sume TTS 1.0 returns

Sume TTS 1.0 takes an output_format with a container of mp3, wav or raw. The default is mp3 at 44.1 kHz and 128 kbps. Sample rate can be 8000, 16000, 22050, 24000, 44100 or 48000. MP3 bit rate runs from 32k to 192k. For wav and raw you choose a PCM encoding: pcm_f32le, pcm_s16le, pcm_mulaw or pcm_alaw.

Getting to Opus from Sume

If you need Opus, ask Sume for 48000 Hz wav and encode with your own tool. Starting from lossless means one lossy step, not two. Do not render to mp3 first and then re-encode, because each lossy generation adds artifacts.

• Set output_format to container wav, sample_rate 48000 and encoding pcm_s16le.
• Encode to Opus in your build step.
• Keep the wav if you still need to join clips, then encode last.

Choosing between them

Pick on the pipeline, not the codec name. A phone system wants 8 kHz mu-law, which both support. A video pipeline wants wav so Timeline audio can join cleanly and a lip-sync model gets lossless input. A web page that just plays a clip is happy with the default mp3.

Which sample rate should you pick

Match the destination. Telephony wants 8000 Hz; speech models for transcription often use 16000 Hz; video work wants 48000 Hz, since that is the rate video containers usually carry. Rendering at the destination rate avoids a resample step.

Raw output is headerless PCM, which suits a pipeline that already knows the rate and encoding. A wav carries a header, which makes it self-describing and safer to hand to other tools.

Takeaway

Sume covers mp3, wav and raw PCM including 8 kHz telephony encodings. Opus is the one ElevenLabs output Sume does not list, and wav at 48 kHz is the clean source for it.