The Workflow was setup to have a clean "GUI" showing only parameters that matter, so you might want to toggle off Link visibility, like in above screenshot.<hr />V1.6 Ace Step 1.5. Turbo and SFT normal and XL model with Ollama. <span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">Text to Audio/Song (examples below<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">)updated the settings for XL models and added a 3rd System Prompt for tags to chose, with more descriptive song descriptions.1.5 XL SFT pipeline now has an "Adaptive Projected Guidance" node and negative prompt.** See below some tips which model and settings to start with.<hr />V1.5 Ace Step 1.5. Turbo and SFT normal and XL model with Ollama:<ul><li>setup to create up to 4 tracks in a run, 2x Ace1.5 and 2x Ace1.5 XL, each with Turbo and SFT model, to compare (can be individually switched on/off)</li><li>VAE changed to tiled Audio VAE decode, uses less Vram.</li></ul><hr />V1.2 Ace Step 1.5. Turbo and SFT model with Ollama Text to Audio/Song<ul><li>small update to GUI, system prompts and SFT sampler "engine"</li></ul><hr />V1.0 Ace Step 1.5. Turbo and SFT model with Ollama Text to Audio/SongAce Step uses TAGS and LYRICS to create a song. These can be generated by Ollama or by own prompts.<ul><li>Can use any Song, Artist as reference or any other description to generate tags and lyrics.</li><li>Will output up to two songs, one generated by Turbo model, the other by the SFT model (experimental).</li><li>Keyscales, bpm and song duration can be randomized.</li><li>able to use dynamic prompts.</li><li>creates suitable songtitle and filenames with Ollama.</li><li>Lora Loader included, hope to see some Loras soon!</li></ul>Avoid sage attention in your comfyui starting parameters, avoid --lowvram setting, as this might force Texencoder to run very slow on CPU instead of GPU.<hr />Download Files:<ul><li>Ace Step 1.5 TURBO model: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files/tree/main/split_files/diffusion_models">https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files/tree/main/split_files/diffusion_models</a></li><li>Ace Step 1.5 SFT model: <a target="_blank" rel="ugc" href="https://huggingface.co/ACE-Step/acestep-v15-sft/tree/main">https://huggingface.co/ACE-Step/acestep-v15-sft/tree/main</a> (download model.safetensor and rename it)</li><li>Ace Step 1.5 XL Models: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files/tree/main/split_files/diffusion_models">https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files/tree/main/split_files/diffusion_models</a>Ace Step 1.5 XL Model merges (i.e. merge of turbo with SFT): <a target="_blank" rel="ugc" href="https://huggingface.co/Aryanne/acestep-v15-test-merges/tree/main/xl">https://huggingface.co/Aryanne/acestep-v15-test-merges/tree/main/xl</a></li><li>Textencoder: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files/tree/main/split_files/text_encoders">https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files/tree/main/split_files/text_encoders </a>(Qwen_0.6b and Qwen_4b required, 1.7b is a smaller alternative to 4b)</li><li>VAE: <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files/tree/main/split_files/vae">https://huggingface.co/Comfy-Org/ace_step_1.5_ComfyUI_files/tree/main/split_files/vae</a></li></ul>Ollama Models, required for tags, lyrics and songtitle, you can choose 1,2 or 3 different models, tags and lyrics might need a bigger model &gt;7b, songtitle can use a smaller model:<ul><li><a target="_blank" rel="ugc" href="https://ollama.com/mirage335/Llama-3-NeuralDaredevil-8B-abliterated-virtuoso">https://ollama.com/mirage335/Llama-3-NeuralDaredevil-8B-abliterated-virtuoso </a>(allround model, fast, usable for tags, lyrics and songtitle, recommended)</li><li><a target="_blank" rel="ugc" href="https://ollama.com/huihui_ai/qwen3-vl-abliterated">https://ollama.com/huihui_ai/qwen3-vl-abliterated </a>(for tags and lyrics, able to use thinking)</li><li><a target="_blank" rel="ugc" href="https://ollama.com/artifish/llama3.2-uncensored">https://ollama.com/artifish/llama3.2-uncensored </a>(small and fast for songtitle and tags)</li></ul><hr />Alternative Turbo and SFT Models (normal, non XL) :<ul><li>Turbo continuous: <a target="_blank" rel="ugc" href="https://huggingface.co/ACE-Step/acestep-v15-turbo-continuous/tree/main">https://huggingface.co/ACE-Step/acestep-v15-turbo-continuous/tree/main</a></li><li>SFT-Shift1: <a target="_blank" rel="ugc" href="https://huggingface.co/ACE-Step/acestep-v15-turbo-shift1/tree/main">https://huggingface.co/ACE-Step/acestep-v15-turbo-shift1/tree/main</a></li><li>SFT-Shift3: <a target="_blank" rel="ugc" href="https://huggingface.co/ACE-Step/acestep-v15-turbo-shift3/tree/main">https://huggingface.co/ACE-Step/acestep-v15-turbo-shift3/tree/main</a></li><li>Merges of SFT, Turbo and Base model: <a target="_blank" rel="ugc" href="https://huggingface.co/Aryanne/acestep-v15-test-merges/tree/main">https://huggingface.co/Aryanne/acestep-v15-test-merges/tree/main</a></li></ul><hr />GGUF Models "normal" and XL: <a target="_blank" rel="ugc" href="https://huggingface.co/Serveurperso/ACE-Step-1.5-GGUF/tree/main">https://huggingface.co/Serveurperso/ACE-Step-1.5-GGUF/tree/main</a><hr /><h3 id="which-models-to-start-with">Which models to start with ?</h3><ul><li>My current choice for normal model: Turbo-SFT merge_ta_0.5 &amp; SFT-Shift1, using these settings:<ul><li>Turbo-SFT_merge model with sampler: er_sde, scheduler: beta57 (or beta), 22 steps</li><li>SFT-Shift1 model with sampler euler, scheduler: normal, 138 steps</li></ul></li><li>XL Model settings:<ul><li>XL Turbo-SFT merge model: sampler: er_sde, scheduler: sgm_uniform, 40 steps<ul><li>alternative: sampler: res_s2, scheduler beta57 (requires RES4LYF custom nodes)</li></ul></li><li>XL SFT model: sampler: euler (or res_2s), scheduler: normal, 46 steps, CFG = 7.3, Adaptive Projected Guidance: eta = 1.05, norm_thresh= 1.3, momentum=0.0. Increase norm_thresh as the main parameter. These settings deliver "stabil" output for XL SFT,Base and their merges. The merges sound way better, pure SFT or Base introduce a lot of noise. I bypassed ModelSamplingAuraflow (see node next to model loader node). I think the base-turbo XL model merge fits well in that slot.</li></ul></li><li>Disable "generate_audio_codes" in "TextEncodeAceStep" node to get different results, it works very well for many genres and reduces process time.</li><li>Ollama Model: Llama-3-NeuralDaredevil-8b-abliterated</li></ul>More infos on models see thread below in discussion.<hr />Save Location:<ul><li>📂 ComfyUI/</li><li>├── 📂 models/</li><li>│ ├── 📂 diffusion_models/</li><li>│ │ └── acestep_v1.5_turbo.safetensors</li><li>│ ├── 📂 text_encoders/</li><li>│ │ ├── qwen_0.6b_ace15.safetensors</li><li>│ │ └── qwen_4b_ace15.safetensors (or 1.7b)</li><li>│ └── 📂 vae/</li><li>│ └── ace_1.5_vae.safetensors</li></ul><hr />Custom Nodes used:<ul><li><a target="_blank" rel="ugc" href="https://github.com/rgthree/rgthree-comfy">https://github.com/rgthree/rgthree-comfy</a></li><li><a target="_blank" rel="ugc" href="https://github.com/yolain/ComfyUI-Easy-Use">https://github.com/yolain/ComfyUI-Easy-Use</a></li><li><a target="_blank" rel="ugc" href="https://github.com/stavsap/comfyui-ollama">https://github.com/stavsap/comfyui-ollama</a></li><li><a target="_blank" rel="ugc" href="https://github.com/KoinnAI/ComfyUI-DynPromptSimplified">https://github.com/KoinnAI/ComfyUI-DynPromptSimplified</a></li></ul>optional (use Beta57 scheduler for a bit more punch, requires RES4LYF): <a target="_blank" rel="ugc" href="https://github.com/ClownsharkBatwing/RES4LYF">https://github.com/ClownsharkBatwing/RES4LYF</a><hr />Examples various styles:<ul><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/firestorms_unleashed_turbo_000">https://soundcloud.com/tele-joe/firestorms_unleashed_turbo_000</a></li><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/westsidevibesforever_funkhipho">https://soundcloud.com/tele-joe/westsidevibesforever_funkhipho</a></li><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/lonelyroadsoul_turbo_00001">https://soundcloud.com/tele-joe/lonelyroadsoul_turbo_00001</a></li><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/hangover">https://soundcloud.com/tele-joe/hangover</a></li><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/sft_secret-paradise">https://soundcloud.com/tele-joe/sft_secret-paradise</a></li><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/cosmicdreamscapesynthpop_turbo">https://soundcloud.com/tele-joe/cosmicdreamscapesynthpop_turbo</a></li><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/mindspin-synthpop_turbo_00001">https://soundcloud.com/tele-joe/mindspin-synthpop_turbo_00001</a></li><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/wildreckoning_thrashmetal_turb">https://soundcloud.com/tele-joe/wildreckoning_thrashmetal_turb</a></li><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/freeflight_metalcore_sft_00001">https://soundcloud.com/tele-joe/freeflight_metalcore_sft_00001</a></li><li><a target="_blank" rel="ugc" href="https://soundcloud.com/tele-joe/urban-ascension_soulful">https://soundcloud.com/tele-joe/urban-ascension_soulful</a></li></ul><hr />Ollama help:<ol><li>Install Ollama from <a target="_blank" rel="ugc" href="https://ollama.com/">https://ollama.com/</a></li><li>download a model: Go to a model page, chose a model , then hit the copy button, i.e. <a target="_blank" rel="ugc" href="https://ollama.com/mirage335/Llama-3-NeuralDaredevil-8B-abliterated-virtuoso">https://ollama.com/mirage335/Llama-3-NeuralDaredevil-8B-abliterated-virtuoso</a></li><li>open terminal and paste the model name, i.e.: ollama run huihui_ai/qwen3-vl-abliterated</li><li>model will be downloaded and can be selected in green comfy node "Ollama Connectivity". Hit "Reconnect" to refresh.</li></ol>

Ace Step 1.5 XL Turbo and SFT - TEXT to AUDIO model with Ollama

Workflow setup to run AceStep 1.5 and 1.5 XL Modelswith updated Settings for XL models

Files with SHA256: dd090fa526ff1fea0fba546f7bea15f3d7b991a2bb2a2d5d494e2e7ea63ac387