> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.ltx.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.ltx.io/_mcp/server.

# Text-to-Audio Workflow

> Generate audio from a text prompt with the LTX-2.5 audio-only ComfyUI workflow, without a video branch.

This guide walks you through the **Text-to-Audio (T2A)** workflow: a ComfyUI workflow that generates audio from a text prompt using LTX-2.5 in audio-only mode. **No video input or output is involved** — the model generates audio directly from your description.

The workflow uses [`LTXVAudioOnlyModel`](/open-source-model/integration-tools/ltx-comfy-ui-nodes#ltxvaudioonlymodel) to disable the model's video pathway, running only the audio stream of the joint audio-video transformer. This produces audio with zero dependence on visual content, while reusing the same checkpoint and text encoder as standard video generation.

## When to Use

Text-to-Audio is the right workflow when you want to generate audio from a text description without any video. If you need audio synchronized to video, use the standard [Text-to-Video](/open-source-model/usage-guides/text-to-video) workflow instead, which generates audio and video jointly.

## Prerequisites

This guide assumes you're familiar with ComfyUI basics. If you're new to LTX in ComfyUI, start with the [ComfyUI setup guide](/open-source-model/integration-tools/comfy-ui).

## Model Files

| File                                                 | Description                                                                   | Placement                          |
| ---------------------------------------------------- | ----------------------------------------------------------------------------- | ---------------------------------- |
| `ltx-2.5-22b-distilled-transformer-bf16.safetensors` | Distilled LTX-2.5 transformer (loaded via **UNETLoader**)                     | `ComfyUI/models/diffusion_models/` |
| `gemma4-12b-with-proj-ltx-2.5-bf16.safetensors`      | Text encoder (Gemma 4 12B)                                                    | `ComfyUI/models/text_encoders/`    |
| `gemma4_e2b_it_bf16.safetensors`                     | Prompt enhancer (Gemma 4 E2B); from `Comfy-Org/gemma-4`, not the LTX-2.5 repo | `ComfyUI/models/text_encoders/`    |
| `ltx-2.5-audio-vae-bf16.safetensors`                 | Audio VAE                                                                     | `ComfyUI/models/vae/`              |

> **Note**
>
> This is an audio-only workflow: it decodes with the audio VAE (`ltx-2.5-audio-vae-bf16.safetensors`) and doesn't load a video VAE.

## Step-by-Step

### 1. Download and Load the Workflow

Download the [Text-to-Audio workflow JSON](https://github.com/Lightricks/ComfyUI-LTXVideo/blob/master/example_workflows/2.5/LTX-2.5_T2A_Single_Stage_Distilled.json) from our GitHub repository and drag it into ComfyUI.

### 2. Install Custom Nodes and Download Models

This workflow requires the [ComfyUI-LTXVideo](https://github.com/Lightricks/ComfyUI-LTXVideo) custom node package. Open the **Workflow Overview** panel (right sidebar) to check for missing nodes or model files.

### 3. Write Your Prompt

Find the positive prompt node and describe the audio you want — the sound, speech, or audio scene. To expand a short prompt, enable the Gemma 4 prompt enhancer (`LTXVGemmaEnhancePrompt`); the negative prompt is not enhanced.

**Example prompt:**

> A woman saying: "Oh, what a lovely day we are having!"

### 4. Set Duration

Set the **number of frames** for your desired audio length; the audio latent length is determined by the frame count and frame rate (audio duration = num-frames ÷ frame-rate). The frame count follows the `1 + a multiple of 8` pattern, the same convention as video.

### 5. Generate

Click **Run**. The 8-step distilled schedule produces audio in seconds.

### 6. Preview

The **PreviewAudio** node at the end of the workflow plays the generated audio directly in ComfyUI.

## How the Pipeline Works

LTX is a joint audio-video transformer that processes both modalities in a single model. The T2A workflow puts it into audio-only mode with [`LTXVAudioOnlyModel`](/open-source-model/integration-tools/ltx-comfy-ui-nodes#ltxvaudioonlymodel), which disables the video stream and cross-modal attention so only the audio pathway runs:

1. **Load model** — The distilled LTX-2.5 transformer is loaded, then `LTXVAudioOnlyModel` disables the video stream and audio↔video cross-attention.
2. **Encode prompt** — Your text is encoded with the Gemma 4 text encoder (the same encoder used for video), with an optional Gemma 4 enhancer to expand short prompts.
3. **Prepare latents** — An empty audio latent is created for the target duration, concatenated with a minimal dummy video latent that the model never processes in audio-only mode.
4. **Sample** — The audio is denoised on the distilled 8-step schedule at CFG 1.
5. **Decode** — The audio latent is separated from the dummy video and decoded through the audio VAE.

## Tips & Troubleshooting

* **Keep cfg near 1.** The distilled model bakes guidance into distillation; explore samplers and cfg but stay near 1.
* **Audio length.** Adjust the frame count to control duration.

## Technical Notes

* LTX is a single joint audio-video transformer that splits its input into `[video, audio]` streams; `LTXVAudioOnlyModel` disables the video stream rather than using a separate audio model.
* A dummy video latent is concatenated with the audio latent via `LTXVConcatAVLatent` before sampling; in audio-only mode those tokens are never attended to.
* After sampling, `LTXVSeparateAVLatent` splits the output and only the audio latent is decoded, via `LTXVAudioVAEDecode`.
* The audio VAE (`ltx-2.5-audio-vae-bf16.safetensors`) is loaded as a separate file via `VAELoader`; no video VAE is loaded in audio-only mode.