upstream-archive/stable-diffusion.cpp
1
0
Fork 0
mirror of https://github.com/leejet/stable-diffusion.cpp.git synced 2026-08-04 17:20:41 -05:00
Code Issues Packages Projects Releases Wiki Activity

feat: add ltx2.3 support (#1463)

Browse Source 
* add GemmaTokenizer

* add basic ltx2.3 support

* change vocab file encoding

* fix ci

* fix ubuntu build

* add temporal tiling support

* add ltx audio support

* update ggml submodule url

* fix generate_video

* add i2v support

* minify bundled Gemma tokenizer vocab sources

* pass video fps into temporal rope embeddings

* fix av_ca_timestep_scale_multiplier

* add LTX2Scheduler support

* update docs

* fix ci
This commit is contained in:
leejet
2026-05-17 16:46:20 +08:00
committed by GitHub
parent 3b4d26f3d9
commit 67dda3f897
44 changed files with 6393 additions and 240 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
examples/common/common.cpp
View File

@@ -340,10 +340,18 @@ ArgOptions SDContextParams::get_options() {
"--high-noise-diffusion-model",
"path to the standalone high noise diffusion model",
&high_noise_diffusion_model_path},
{"",
"--embeddings-connectors",
"path to LTXAV embeddings connectors",
&embeddings_connectors_path},
{"",
"--vae",
"path to standalone vae model",
&vae_path},
{"",
"--audio-vae",
"path to standalone LTX audio vae model",
&audio_vae_path},
{"",
"--taesd",
"path to taesd. Using Tiny AutoEncoder for fast decoding (low quality)",
@@ -669,7 +677,9 @@ std::string SDContextParams::to_string() const {
<< " llm_vision_path: \"" << llm_vision_path << "\",\n"
<< " diffusion_model_path: \"" << diffusion_model_path << "\",\n"
<< " high_noise_diffusion_model_path: \"" << high_noise_diffusion_model_path << "\",\n"
<< " embeddings_connectors_path: \"" << embeddings_connectors_path << "\",\n"
<< " vae_path: \"" << vae_path << "\",\n"
<< " audio_vae_path: \"" << audio_vae_path << "\",\n"
<< " taesd_path: \"" << taesd_path << "\",\n"
<< " esrgan_path: \"" << esrgan_path << "\",\n"
<< " control_net_path: \"" << control_net_path << "\",\n"
@@ -728,7 +738,9 @@ sd_ctx_params_t SDContextParams::to_sd_ctx_params_t(bool vae_decode_only, bool f
llm_vision_path.c_str(),
diffusion_model_path.c_str(),
high_noise_diffusion_model_path.c_str(),
embeddings_connectors_path.c_str(),
vae_path.c_str(),
audio_vae_path.c_str(),
taesd_path.c_str(),
control_net_path.c_str(),
embedding_vec.data(),
@@ -821,7 +833,7 @@ ArgOptions SDGenerationParams::get_options() {
&hires_upscaler},
{"",
"--extra-sample-args",
"extra sampler args, key=value list. Currently lcm supports noise_clip_std, noise_scale_start, noise_scale_end",
"extra sampler/scheduler args, key=value list. lcm supports noise_clip_std, noise_scale_start, noise_scale_end; ltx2 supports max_shift, base_shift, stretch, terminal",
&extra_sample_args},
};
@@ -1006,6 +1018,11 @@ ArgOptions SDGenerationParams::get_options() {
"process vae in tiles to reduce memory usage",
true,
&vae_tiling_params.enabled},
{"",
"--temporal-tiling",
"enable temporal tiling for LTX video VAE decode",
true,
&vae_tiling_params.temporal_tiling},
{"",
"--hires",
"enable highres fix",
@@ -1270,7 +1287,7 @@ ArgOptions SDGenerationParams::get_options() {
on_high_noise_sample_method_arg},
{"",
"--scheduler",
"denoiser sigma scheduler, one of [discrete, karras, exponential, ays, gits, smoothstep, sgm_uniform, simple, kl_optimal, lcm, bong_tangent], default: discrete",
"denoiser sigma scheduler, one of [discrete, karras, exponential, ays, gits, smoothstep, sgm_uniform, simple, kl_optimal, lcm, bong_tangent, ltx2], default: model-specific",
on_scheduler_arg},
{"",
"--sigmas",
@@ -1703,6 +1720,9 @@ bool SDGenerationParams::from_json_str(
if (tiling_json.contains("enabled") && tiling_json["enabled"].is_boolean()) {
vae_tiling_params.enabled = tiling_json["enabled"];
}
if (tiling_json.contains("temporal_tiling") && tiling_json["temporal_tiling"].is_boolean()) {
vae_tiling_params.temporal_tiling = tiling_json["temporal_tiling"];
}
if (tiling_json.contains("tile_size_x") && tiling_json["tile_size_x"].is_number_integer()) {
vae_tiling_params.tile_size_x = tiling_json["tile_size_x"];
}
@@ -2212,6 +2232,7 @@ sd_vid_gen_params_t SDGenerationParams::to_sd_vid_gen_params_t() {
params.strength = strength;
params.seed = seed;
params.video_frames = video_frames;
params.fps = fps;
params.vace_strength = vace_strength;
params.vae_tiling_params = vae_tiling_params;
params.cache = cache_params;
@@ -2300,6 +2321,7 @@ std::string SDGenerationParams::to_string() const {
<< ", upscale_tile_size: " << hires_upscale_tile_size << " },\n"
<< " vae_tiling_params: { "
<< vae_tiling_params.enabled << ", "
<< vae_tiling_params.temporal_tiling << ", "
<< vae_tiling_params.tile_size_x << ", "
<< vae_tiling_params.tile_size_y << ", "
<< vae_tiling_params.target_overlap << ", "