enne2/qwentts.cpp
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

server: sampling overrides on the speech endpoint

Browse Source 
The speech body accepts seed, max_new_tokens, temperature, top_k,
top_p, and repetition_penalty. Unset fields keep the engine defaults,
a temperature of zero selects greedy decoding, and the subtalker
mirrors the talker knobs. A fixed seed makes a request reproducible.
This commit is contained in:
Pascal
2026-07-05 15:01:37 +02:00
parent 1a680e8816
commit 73fe0c67bb
4 changed files with 100 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/ARCHITECTURE.md
+7 -1
View File
@@ -626,7 +626,13 @@ Endpoints :
```
POST /v1/audio/speech OAI text-to-speech; response_format "pcm"
streams s16le 24 kHz mono chunked as it is
generated, "wav" returns a one-shot RIFF file
generated, "wav" returns a one-shot RIFF file.
Optional sampling overrides ride in the same
body: seed, max_new_tokens, temperature,
top_k, top_p, repetition_penalty. Unset
fields keep the engine defaults, temperature
0 selects greedy decoding, the subtalker
mirrors the talker knobs
GET /v1/models single loaded model
GET /v1/voices model speakers plus registered cloned voices
POST /v1/voices register a cloned voice: {name, ref_text,