How to Use Seed Audio 2.0 Online

Start in the live audio workspace on the homepage. You can read the controls before signing in, but you need an account to submit a generation request. Keep this guide open in another tab if you want to follow along.

1. Select Open audio generator below to jump directly to the workspace. You do not need to install a desktop app.
2. Use Get Started Free in the navigation to open the account dialog. Create an account or use the sign-in option if you already have one.
3. After signing in, check the credit balance at the top right. Open the pricing page if you need to review credit packs before purchasing.

The workspace displays a usage rate of 5 credits per minute and supports up to 2 minutes per generation. Review the current pricing information before generating so you know which credit pack fits your intended usage.

Find AI model at the top of the workspace. Seed Audio 1.0 is ready to use now, and Seed Audio 2.0 is coming soon. Open the dropdown to explore the model options, then choose Seed Audio 1.0 to start creating.

1. Open the AI model dropdown at the top of the tool.
2. Select Seed Audio 1.0 to generate dialogue, sound effects, and audio scenes with the currently available model.
3. Seed Audio 2.0 is labeled Coming soon. You can select it, but Generate audio stays disabled until you switch back to Seed Audio 1.0.

Seed Audio 2.0 is coming soon. In the meantime, bring your ideas to life with Seed Audio 1.0: choose a voice or reference, describe your scene, and generate your audio.

Under Voice & references, choose one of four modes. For a first text-to-audio test, select Prompt only. It lets you hear how your written scene performs without introducing a selected voice or uploaded reference.

1. Choose Prompt only to generate audio from your description without sending a voice selection, audio reference, or image reference.
2. Choose Voice library to browse built-in voices. Search by name, accent, or gender, then select a voice. Auto-select voice lets the model choose instead. The full voice library also has Use this voice links that open the generator with a voice selected.
3. Choose Reference audio to click or drag files into the upload box. Wait for uploading to finish, then listen to the reference player below the link field. Remove a clip with its remove control if necessary.
4. Choose Reference image to click or drag one image into the upload box. The uploaded image appears inside that box. Click it or drop another file there to replace it.

The modes are mutually exclusive. A saved voice or reference may still be there when you switch back, but only the current mode is sent when you generate. Use reference material you have permission to upload.

In Describe your audio, explain what the listener should hear. The input supports up to 2,800 characters, with a counter in the lower-right corner. Start with one clear situation rather than a long list of unrelated sounds.

1. Describe the setting and the main sound: a host speaking in a quiet studio, two travelers on a platform, or rain falling beside a stream.
2. For speech, name each speaker and include the exact words in quotation marks. Specify the language, emotion, and delivery in plain English, such as a calm voice at a conversational pace.
3. Add ambience, music, and sound effects only when they help the scene. Describe which sound should stay in the foreground and what should remain quiet behind it.
4. Put events in listening order. Say whether a sound happens before, during, or after a line. If you selected a reference, explain the voice quality or scene mood it should guide.

You do not need to specify a duration. Let the dialogue and sound events determine the length. If you do request a duration, treat it as creative direction rather than an exact editing boundary.

For a first attempt, keep the default settings. Open Show settings only if you need a different output format, sample rate, or voice adjustment. These controls change the request; they do not replace a clear prompt.

1. Under Audio output, File format defaults to MP3. This is a convenient starting point for listening and sharing. The other output options are WAV, PCM, and Opus.
2. Leave Sample rate at its default of 24 kHz unless your editing workflow needs another available rate. The selector offers options from 8 kHz through 48 kHz.
3. Under Voice controls, Speed changes the requested pace, Volume changes the requested level, and Pitch adjusts the voice in semitones. Start with their defaults and change one control at a time.
4. Use Reset settings if you want to return the output and voice controls to their defaults. It does not clear your prompt or switch your reference mode.

Output format and upload format are different settings. PCM is an available output option, but raw PCM reference uploads are not supported. If you are unsure which output to use, leave MP3 selected.

Review the prompt and selected mode, then select Generate audio. If you are not signed in, complete the sign-in dialog. Reference audio and image modes need a successful upload or an added link before you can use them.

1. Make sure Describe your audio is not empty and the character counter is within 2,800. Finish any file upload before submitting.
2. Select Generate audio once. The right-hand Audio preview panel changes to GENERATING and displays a loading indicator and progress bar.
3. Your audio may take anywhere from around 15 seconds to a few minutes to generate, depending on the length and complexity of your scene.

If generation fails, read the error before submitting another request. Check the troubleshooting section below. If you return to the site later, look in My Creations before starting the same scene again.

When generation succeeds, Audio preview changes to READY and displays your result. The initial SAMPLE player is a demonstration, so check the status before assuming that a clip is your new generation.

1. Press Play in Audio preview. If your browser does not start playback automatically, use the play button or the audio player controls.
2. Listen for the words you supplied, the distinction between speakers, the balance of speech and background sound, and the order of important effects.
3. Open My Creations and locate the completed item. Use its Download button to save a copy. This is also the place to look if your browser’s preview player does not show a download option.

Keep the downloaded file together with the prompt and settings you used. A request to generate the same scene again can produce a different result; saving a file preserves the version you chose.

If the first result needs work, identify one audible issue and change the instruction responsible for it. A focused revision makes it easier to understand what helped, instead of changing the voice, music, ambience, and wording all at once.

1. For hard-to-hear dialogue, keep the script and ask for quieter music or ambience. For an unclear speaker change, label the speakers and avoid overlapping lines.
2. For a crowded scene, remove an unnecessary sound effect or use one main ambience. For timing problems, describe the sequence as first, then, and finally.
3. Generate a new version and compare it with the saved one. Each new generation is a separate request; check your credits as you experiment.

You can revise the prompt without re-uploading a reference that is already present. Before generating, confirm that the intended mode is still selected, especially if you switched between Prompt only and a reference mode.