audio-cpp commited on
Commit
fd2fe4c
·
verified ·
1 Parent(s): 9251401

Upload folder using huggingface_hub

Browse files
.gitattributes CHANGED
@@ -260,3 +260,9 @@ Haas-ja-v1.0-GGUF/haas-ja-v1.0-q8_0.gguf filter=lfs diff=lfs merge=lfs -text
260
  Underdog-Bark-3.0-GGUF/underdog-bark-3.0-mixed-f16.gguf filter=lfs diff=lfs merge=lfs -text
261
  MuScriptor-Small-GGUF/muscriptor-large-f32.gguf filter=lfs diff=lfs merge=lfs -text
262
  MuScriptor-Small-GGUF/muscriptor-medium-f32.gguf filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
260
  Underdog-Bark-3.0-GGUF/underdog-bark-3.0-mixed-f16.gguf filter=lfs diff=lfs merge=lfs -text
261
  MuScriptor-Small-GGUF/muscriptor-large-f32.gguf filter=lfs diff=lfs merge=lfs -text
262
  MuScriptor-Small-GGUF/muscriptor-medium-f32.gguf filter=lfs diff=lfs merge=lfs -text
263
+ PocketTTS-Small-GGUF/english-18m/english-18m-f32.gguf filter=lfs diff=lfs merge=lfs -text
264
+ PocketTTS-Small-GGUF/english-18m-refl/english-18m-refl-f32.gguf filter=lfs diff=lfs merge=lfs -text
265
+ PocketTTS-Small-GGUF/english-2m/english-2m-f32.gguf filter=lfs diff=lfs merge=lfs -text
266
+ PocketTTS-Small-GGUF/english-2m-refl/english-2m-refl-f32.gguf filter=lfs diff=lfs merge=lfs -text
267
+ PocketTTS-Small-GGUF/english-7m/english-7m-f32.gguf filter=lfs diff=lfs merge=lfs -text
268
+ PocketTTS-Small-GGUF/english-7m-refl/english-7m-refl-f32.gguf filter=lfs diff=lfs merge=lfs -text
PocketTTS-Small-GGUF/README.md ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ base_model:
4
+ - kyutai/pocket-tts
5
+ library_name: audio.cpp
6
+ language:
7
+ - en
8
+ tags:
9
+ - audio.cpp
10
+ - gguf
11
+ - text-to-speech
12
+ - voice-cloning
13
+ - streaming
14
+ pipeline_tag: text-to-speech
15
+ ---
16
+
17
+ # PocketTTS Small GGUF
18
+
19
+ Follow or star [audio.cpp](https://github.com/0xShug0/audio.cpp) for the inference code and updates.
20
+
21
+ English PocketTTS Small checkpoints from [Kyutai](https://huggingface.co/kyutai/pocket-tts/tree/db04f0e653473e5c7333c69686ac92b7460b0c1d/languages), converted to F32 GGUF. The `pocket_tts_small` family supports text-to-speech, reference-audio voice cloning, and streaming.
22
+
23
+ | Variant | Weights | Predefined voice |
24
+ |---|---|---|
25
+ | 18M ReFL (default) | `english-18m-refl/english-18m-refl-f32.gguf` | `english-18m-refl/embeddings/alba.safetensors` |
26
+ | 18M | `english-18m/english-18m-f32.gguf` | `english-18m/embeddings/alba.safetensors` |
27
+ | 7M ReFL | `english-7m-refl/english-7m-refl-f32.gguf` | `english-7m-refl/embeddings/alba.safetensors` |
28
+ | 7M | `english-7m/english-7m-f32.gguf` | `english-7m/embeddings/alba.safetensors` |
29
+ | 2M ReFL | `english-2m-refl/english-2m-refl-f32.gguf` | `english-2m-refl/embeddings/alba.safetensors` |
30
+ | 2M | `english-2m/english-2m-f32.gguf` | `english-2m/embeddings/alba.safetensors` |
31
+
32
+ Keep each variant's own `embeddings` directory beside its GGUF. The configuration and tokenizer are embedded in the GGUF. Voice states belong to their respective checkpoints and should not be exchanged between variants.
33
+
34
+ ## Usage
35
+
36
+ ```sh
37
+ audiocpp_cli --family pocket_tts_small --task tts \
38
+ --model PocketTTS-Small-GGUF/english-18m-refl/english-18m-refl-f32.gguf \
39
+ --backend cpu --voice-id alba \
40
+ --text "Hello! This is the small PocketTTS model speaking locally." \
41
+ --out output.wav --log
42
+ ```
43
+
44
+ Select `--mode streaming` for streaming synthesis. To clone a voice, use `--task clon --voice-ref reference.wav` instead of `--voice-id alba`. GPU backends use GGML graphs; select `--backend cuda` or `--backend vulkan` in a corresponding build.
45
+
46
+ Offline waveform decoding uses bounded causal blocks to limit memory usage. Use `--session-option pocket_tts_small.attention=flash` and `--session-option pocket_tts_small.kv_cache_type=f16` to select flash acoustic attention and F16 KV storage independently. Defaults are eager attention and F32 KV storage; the GGUF weights remain unchanged. See the [model documentation](https://github.com/0xShug0/audio.cpp/blob/main/docs/models/pocket_tts_small.md) for controls and the experimental C++ batch interface.
47
+
48
+ ## Attribution
49
+
50
+ Original models and implementation: [Kyutai PocketTTS](https://github.com/kyutai-labs/pocket-tts), code revision [`4cc57533`](https://github.com/kyutai-labs/pocket-tts/tree/4cc5753319ceddc5cbf70598d9011f1f22858ba0). Checkpoint revision: [`db04f0e6`](https://huggingface.co/kyutai/pocket-tts/tree/db04f0e653473e5c7333c69686ac92b7460b0c1d). Model weights are licensed under CC-BY-4.0.
PocketTTS-Small-GGUF/english-18m-refl/embeddings/alba.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e1aa12e17245952e044853867deef5e77ec4474df119cbeefa4ed2a3d878a7b1
3
+ size 4163992
PocketTTS-Small-GGUF/english-18m-refl/english-18m-refl-f32.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f19bbf172c443ce63aaf22ce2d2fbabf45726cc48694c3c4904e8ec4a8661a3c
3
+ size 111809536
PocketTTS-Small-GGUF/english-18m/embeddings/alba.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11554cbe9f3d3a3e10722932df9952bb44928e534f7af3200524659166b5627a
3
+ size 4163992
PocketTTS-Small-GGUF/english-18m/english-18m-f32.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d66ba8fa4cce2c12901a46a93ead2785ea5e6ac689e6127ffec14c41c17b8031
3
+ size 111809536
PocketTTS-Small-GGUF/english-2m-refl/embeddings/alba.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac4d7735dc6b5d0591400660a23e9eace974aa1f4085a5be4a8ea2ddfc4ecce8
3
+ size 522632
PocketTTS-Small-GGUF/english-2m-refl/english-2m-refl-f32.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aaa47812c03c0de62811fe4abbaaa43718741e911e7dd20bf25f190a0065a5e5
3
+ size 46792352
PocketTTS-Small-GGUF/english-2m/embeddings/alba.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f561ebfdb6670e89b441bcdf35b3e86e25a620dfaa4eaf5637a63833d367b7c
3
+ size 522632
PocketTTS-Small-GGUF/english-2m/english-2m-f32.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:107e012237171ca2aa4c940818da7db34fc25657eeec95cf5e2db5830a426614
3
+ size 46792352
PocketTTS-Small-GGUF/english-7m-refl/embeddings/alba.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2132e156c08f57779a694bb4b57174b1dd8dd1ccb414e8d8df090a648fac13cd
3
+ size 1042832
PocketTTS-Small-GGUF/english-7m-refl/english-7m-refl-f32.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43aabac6b8778d7ae2d41cf5a22068119cdf05e80710815510a2b188ead82913
3
+ size 66564096
PocketTTS-Small-GGUF/english-7m/embeddings/alba.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fef27ae7a34132e88c81090316145671362da386ab4e5c6b048e5d16800d746b
3
+ size 1042832
PocketTTS-Small-GGUF/english-7m/english-7m-f32.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e58b4a0643b5b9b61b6bd1c9caf1bfb1fb0f457c46534d32cd5ff798296cb2e1
3
+ size 66564096