Historical professional tier for lyric-to-song creation and iteration of existing works
FUZZ-1.0 Pro is the historical Pro-tier music model in the Riffusion FUZZ series, designed for text-based concepts, custom lyric-to-song creation, and continued creation from existing songs. It is suitable for music projects that need to explicitly retain model selection, and can also use operations supported by this model to continue creating around a song ID, organizing the creative process into audio tasks that can be previewed and modified.
Input parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
Specifications and interface features
Clarify capacity, inputs and outputs, and invocation methods before selecting a model.
Model positioning
Historical Pro tier of the FUZZ series; invocation value: FUZZ-1.0 Pro
Creation inputs
Text prompts, custom lyrics, song title
Creation modes
Inspiration generation, custom lyric generation; the interface provides an instrumental toggle
Work iteration
Select from action operations supported by this model; validate separately from other tiers
Time control
continue_at specifies the continuation position; clip replacement specifies start and end positions in seconds
Result delivery
JSON results, including audio_url, song ID, lyrics, status, and duration
Task method
Supports asynchronous task ID queries and callbacks
The historical Pro tier is a model positioning; the above creation and delivery specifications follow the corresponding invocation interface documentation on this platform and do not represent native capacity metrics.
Core Capabilities
Learn what FUZZ-1.0 Pro can bring to your work.
Express lyrics and sound direction separately
When customizing a song with /riffusion/audios, you can set custom=true, put the lyrics in lyric, and use style to describe the genre, instruments, and mood; inspiration mode, which starts from a textual concept, uses prompt. When using /producer/audios, music prompts are submitted through prompt, and requests must include action, lyric, and prompt. This input organization is suitable for song drafts that already have lyrics but still need to explore vocal and arrangement directions.
Continue creating from an existing song
FUZZ-1.0 Pro iterations can revolve around audio_id: covers are used to try new song expressions, while continuations use continue_at to specify where to continue. Creators can first listen to an existing version, then decide the direction of the next revision, without having to start from a completely independent new song every time.
Focus edits on a specified segment
When an issue is concentrated in a section of lyrics or a transition point, you can use replace_section, define the editing range with start and end times, and submit new lyrics. It is suitable for turning revision feedback into a specific interval; after completion, listen together with the preceding and following sections to check whether the transition and song narrative feel natural.
Use Cases
Start with specific tasks to find where the model can be effective.
Listening demos for original lyrics
Enter segmented lyrics, a title, and the desired musical direction to generate song demos for the creative team to review. It is recommended to first define the verse and chorus content, then describe the instruments, mood, and vocal feel. Deliver the audio link and lyric record to facilitate discussion of melodic expression rather than remaining only at the text proposal stage.
Cover exploration for the same work
After you have an existing song, save its audio_id, then initiate a cover task with cover, combining lyrics and style descriptions to explore new expressions. This is suitable for comparing different sound directions during the song proposal stage; listen to the original and new versions together, focusing on whether the mood, lyric expression, and overall style meet the project requirements.
Adding sections and local revisions
For works that need additional content, prepare continuation lyrics and specify where to continue; for works that need changes only to one section, submit replacement lyrics and a time range. After outputting the new audio result, conduct a full listening review and archive the version, making it suitable for turning feedback such as “add a section” or “change this section” into specific tasks.
How to choose this model
Choose based on task complexity, input materials, and expected results.
When to choose FUZZ-1.0 Pro
If an existing project explicitly uses FUZZ-1.0 Pro, or if you want to use this legacy professional tier as a fixed audition target, you can continue selecting it and record the lyrics, prompt, and song ID. FUZZ-1.0 and FUZZ-1.0 Pro are different options and cannot replace each other simply because their names are similar; the Pro name also does not mean a fixed degree of audio quality improvement.
How to weigh it against later versions
FUZZ-1.1 Pro and FUZZ-2.0 Pro are different models, not compatible aliases for FUZZ-1.0 Pro. For new projects, you can use the same lyrics and musical direction for audition comparisons, then choose based on the result. If the main requirement is the editing tools or export workflow of newer versions, select the corresponding feature separately; do not directly treat the capabilities of later versions as capabilities of this model.
Getting started: test one local revision in an existing creation
Prepare the input first, then proceed to the corresponding application workflow.
Prepare input
Prepare a usable song ID, the segment to be modified, and the musical direction you want to retain.
Organize the request and subsequent workflow
Submit the creation using the exact FUZZ-1.0 Pro invocation value; select the prompt, lyric, and title fields for the corresponding entry point, and save the task_id and work ID. Complete the audition first, then choose subsequent actions according to this model's documentation, without mixing in editing fields from other tiers.
Practical task example: test one local revision in an existing creation
Design the task directly from the following inputs and acceptance priorities.
Suggested task
Generate a draft first, then use audio_id and the segment start and end times according to the operations for this model to try modifying one clearly defined section.
Key checks
Check the transition between the modified section and the surrounding sections, and retain the original track for comparison; shared action enumerations do not mean that all editing actions apply to this model.
Usage Boundaries
Before formal use, understand the output quality and capability boundaries.
Continuation and segment replacement are forms of generative creation and should not be delivered as precise, audio-scissor-style edits. Specified times indicate where processing should occur; after completion, you still need to check the joins, replacement segment, and surrounding content, especially by listening to lyric transitions and the overall musical direction.
The parameters for the two endpoints cannot be directly exchanged: /riffusion/audios provides style, while /producer/audios uses its own prompt and control fields. When migrating existing requests, keep the model name unchanged while adjusting fields and action names for the target endpoint.
The audio_url in task results is the audio delivery endpoint. Do not treat the nullable video_url as a guaranteed music video, and do not treat the duration of a particular result as a fixed song length. If video or a specific audio format is needed, arrange separate media-processing steps.
Frequently Asked Questions
Answers to common questions when using FUZZ-1.0 Pro.
Are FUZZ-1.0 Pro and FUZZ-1.0 the same model?
No. They are separately listed models, and FUZZ-1.0 Pro belongs to the historical Pro tier. When calling it, explicitly specify FUZZ-1.0 Pro to avoid relying on the default selection. When comparing the two, it is recommended to listen using the same lyrics and musical direction rather than judging the final result solely by the Pro name.
Is a single musical idea enough, or must I prepare complete lyrics?
Creation can start from a text concept, or you can use custom lyrics. In the request structure, /riffusion/audios requires action and lyric, while /producer/audios also requires prompt; inspiration mode must retain required fields as well, so you cannot directly copy simplified requests that omit fields.
How do I continue a song generated by FUZZ-1.0 Pro?
Save the song result's id, use it as audio_id in the continuation request, set action to extend, and use continue_at to specify the position from which to continue, in seconds. Also prepare continuation lyrics or a musical direction, then listen to the transition between the old and new content after completion.
Can I modify only one section of lyrics in a song?
You can use the replace_section workflow, submitting the song's audio_id, the replacement lyrics, and the time range in seconds specified by replace_section_start and replace_section_end. This regenerates the specified segment rather than replacing an existing recording word by word, so check the connections before and after the change.
Can a generation task be completed in the background and retrieved later?
Yes. After setting async=true, first obtain task_id, then query the task status; you can also configure callback_url to receive the completed result. The application should save the task ID, read audio_url in data once the result is ready, and save the song ID for later covers, continuations, or partial modifications.
