llama.cpp
mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary)
#26254

Merged

mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary) #26254

ngxson merged 56 commits into master from xsn/qwen3-tts

convert text model

ae7663e8

main model load ok

8892b6c6

convert encoder ok

de0ac58c

ngxson changed the title ~~convert text model~~ mtmd: support Qwen-TTS 20 days ago

ngxson changed the title ~~mtmd: support Qwen-TTS~~ mtmd: support Qwen3-TTS 20 days ago

github-actions added conversion

speaker encoder loading ok

39b9c2d5

speaker enc graph

4da8d180

github-actions added mtmd

Merge branch 'master' into xsn/qwen3-tts

b98aff26

adapt vocab for backbone (with some tricks)

894ac82b

Merge branch 'master' into xsn/qwen3-tts

ebd17cc1

add suppress_tokens

85af600c

github-actions added model

poc new mtmd gen api

eea7c8b3

convert code_predictor to gguf

ed0b0503

load gen_code model ok

bd38cb5c

add clip_encode

4712dce7

wire up

2e9e8bbe

code gen cgraph init version

d98264a5

ngxson force pushed from 7afa5b05 to d98264a5 18 days ago

Merge branch 'master' into xsn/qwen3-tts

e68884dc

code2wav convert to gguf

bf0e641f

code2wav graph ok

ff6b2892

wire up in/out

84559b46

(wip) subgraph

79a99219

wire up

13bfdc66

wip, correct code2wav

51027785

ngxson force pushed from 55467b6b to 51027785 17 days ago

demo (to be removed)

2348de10

code2wav preserve kv between calls

a9df03d0

demo voice clone

3e9e4849

llama: add llama_model_get_tok_embd

1f096f4a

mtmd_helper_gen_audio API

ffe530be

github-actions added documentation

fix clamp cold prefix

914fd950

fuse snake op

06aae588

demo: use proper sampling

eae9c2c3

update dev docs

86020c46

polymorphism helper

22bc9df3

revamp llama-tts binary

8f94fc52

update docs

bd52f360

fix compile

82dc811a

ngxson changed the title ~~mtmd: support Qwen3-TTS~~ mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary) 16 days ago

fix lint

511cc2fa

nits

bfc0714a

add guide + docs

4523db99

more timings info

8a402097

ngxson marked this pull request as ready for review 16 days ago

ngxson requested a review from

ggerganov 16 days ago

ngxson requested a review 16 days ago

ngxson requested a review from

CISC 16 days ago

ngxson requested a review 16 days ago

ngxson commented on 2026-08-01

clean up code comments

e298a2e4

Merge branch 'master' into xsn/qwen3-tts

f695c222

security fixes

4a9e349f

update docs

026d35cf

use ggml_build_forward_select, clean up comments

2307bf5a

fix ci

dc5dd58b

Merge branch 'master' into xsn/qwen3-tts

6e4da4a7

ngxson requested a review from

JohannesGaessler 16 days ago

github-actions added testing

ngxson removed review request from

JohannesGaessler 16 days ago

ggerganov assigned

ggerganov 15 days ago

ggerganov approved these changes on 2026-08-02

Merge branch 'master' into xsn/qwen3-tts

0b6e3402

use ISO 639-1 language code

1df59d38

rename CODE2WAV --> GEN_WAV, update docs

d827cded

clean up

2f240361

clean up tts.cpp

127db3f0

ngxson commented on 2026-08-02

add seq_id

ee1a7295

add step_prompt()

2690f91b

mtmd_helper_model_can_chat

564f25d7

Merge branch 'master' into xsn/qwen3-tts

2f40d38c

CISC approved these changes on 2026-08-04

clean up comments

201e50cc

ngxson merged 07132750 into master 13 days ago

ngxson commented on 2026-08-04

Reviewers

CISC

ggerganov

Assignees

ggerganov

Labels

documentation model testing mtmd conversion

Milestone

No milestone

llama.cpp mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary) #26254 Merged

mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary) #26254

llama.cpp
mtmd: support Qwen3-TTS (note: breaking change to llama-tts binary)
#26254

Merged