cinemasetfree.com
← Back to all generators
lucataco/magnet
Official
View on Replicate →
MAGNeT: Masked Audio Generation using a Single Non-Autoregressive Transformer
Capabilities
Top-P
Cost
Community model (estimated from hardware time)
Input Parameters
| Name | Type | Description | Default | Constraints |
|---|---|---|---|---|
decoding_steps_stage_1 | integer | Number of decoding steps for stage 1 | 20 | — |
decoding_steps_stage_2 | integer | Number of decoding steps for stage 2 | 10 | — |
decoding_steps_stage_3 | integer | Number of decoding steps for stage 3 | 10 | — |
decoding_steps_stage_4 | integer | Number of decoding steps for stage 4 | 10 | — |
max_cfg | number | Max CFG coefficient | 10 | — |
min_cfg | number | Min CFG coefficient | 1 | — |
model | string | Model to use | "facebook/magnet-small-10secs" | facebook/magnet-small-10secs facebook/magnet-medium-10secs facebook/magnet-small-30secs facebook/magnet-medium-30secs facebook/audio-magnet-small facebook/audio-magnet-medium |
prompt | string | Input Text | "80s electronic track with melodic synthesizers, catchy beat and groovy bass" | — |
span_score | string | An enumeration. | "prod-stride1" | max-nonoverlap prod-stride1 |
temperature | number | Temperature for sampling | 3 | — |
top_p | number | Top p for sampling | 0.9 | min: 0, max: 1 |
variations | integer | Number of variations to generate | 3 | min: 1, max: 4 |
decoding_steps_stage_1 integer Number of decoding steps for stage 1
Default:
20 decoding_steps_stage_2 integer Number of decoding steps for stage 2
Default:
10 decoding_steps_stage_3 integer Number of decoding steps for stage 3
Default:
10 decoding_steps_stage_4 integer Number of decoding steps for stage 4
Default:
10 max_cfg number Max CFG coefficient
Default:
10 min_cfg number Min CFG coefficient
Default:
1 model string Model to use
Default:
"facebook/magnet-small-10secs" facebook/magnet-small-10secs facebook/magnet-medium-10secs facebook/magnet-small-30secs facebook/magnet-medium-30secs facebook/audio-magnet-small facebook/audio-magnet-medium
prompt string Input Text
Default:
"80s electronic track with melodic synthesizers, catchy beat and groovy bass" span_score string An enumeration.
Default:
"prod-stride1" max-nonoverlap prod-stride1
temperature number Temperature for sampling
Default:
3 top_p number Top p for sampling
Default:
0.9 min: 0, max: 1 variations integer Number of variations to generate
Default:
3 min: 1, max: 4 Version:
e8e2ecd4a1da Updated: 2/26/2026 2.9K runs