Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Modulate / Velma-2

Audio Event Detection

Whole-file detection of 42 non-speech sound events, returning only labels that clear the model’s confidence threshold.

Audio processing
availableClosed weightsAudio processing
Official documentation (opens in a new tab)Reviewed 13 Oct 2026
7 of 20 specification fields collected
Context windowNot collectedAwaiting source review
Maximum outputNot collectedAwaiting source review
Input billingSee ratesModulate Models API · batch
Output billingSee ratesPublished units and conditions below

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
Not collected
Maximum output
Not collected
Input modalities
AudioSource for Input modalities (opens in a new tab)
Output modalities
Classification metadataSource for Output modalities (opens in a new tab)
Reasoning
Not collected
Tool calling
Not collected
Structured output
Not collected

Fixed response JSON fields are documented. No caller-supplied constrained generation schema has been verified.

Streaming
Not collected

Only a batch endpoint is documented for this family. Unpublished streaming access is not inferred from broad marketing copy.

Batch processing
SupportedSource for Batch processing (opens in a new tab)

Synchronous file-processing endpoint; this does not imply asynchronous multi-file jobs or a discounted text batch API.

Fine-tuning
Not collected
Knowledge cutoff
Not collected
Release date
Not collected
Parameters
Not collected
Architecture
Not collected
Languages
Not collected
Context caching
Not collected

No discounted prompt-cache or persistent model-cache contract verified for these audio endpoints. Streaming buffers, stored clips and per-session speaker labels are not prompt caching.

Maximum audio file
100 MB per prerecorded uploadSource for Maximum audio file (opens in a new tab)
Billing conditions
Per audio hour, with separate batch/live rates where published; see endpoint rates for optional enrichment chargesSource for Billing conditions (opens in a new tab)
Classification labels
View all 42 entries
  • Acoustic_guitar
  • Applause
  • Bark
  • Bass_drum
  • Burping_or_eructation
  • Bus
  • Cello
  • Chime
  • Clarinet
  • Computer_keyboard
  • Cough
  • Cowbell
  • Crying
  • Double_bass
  • Drawer_open_or_close
  • Electric_piano
  • Fart
  • Finger_snapping
  • Fireworks
  • Flute
  • Glockenspiel
  • Gong
  • Gunshot_or_gunfire
  • Harmonica
  • Hi_hat
  • Keys_jangling
  • Knock
  • Laughter
  • Meow
  • Microwave_oven
  • Oboe
  • Saxophone
  • Scissors
  • Shatter
  • Snare_drum
  • Squeak
  • Tambourine
  • Tearing
  • Telephone
  • Trumpet
  • Violin_or_fiddle
  • Writing
Source for Classification labels (opens in a new tab)
Output formats
JSON probs mapping for confident events plus duration_msSource for Output formats (opens in a new tab)

Keys are sparse; {} is valid. Missing keys are not measured zero probability. Event scores are independent and need not sum to one.

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

Modulate Models API · batch

Provider source (opens in a new tab)
API model ID
/api/velma-2-audio-event-classifier

Endpoint path selects the served family; not a request model parameter or immutable checkpoint ID.

Access
POST https://platform.modulate.ai/api/velma-2-audio-event-classifier with X-API-Key and multipart upload_file. Response is synchronous.
Regions
Public endpoint deployment region and regional routing are not specified in the reviewed API contract.
Input modalities
AudioSource for Input modalities (opens in a new tab)
Output modalities
Classification metadataSource for Output modalities (opens in a new tab)
Streaming
Not supportedSource for Streaming (opens in a new tab)
Batch processing
SupportedSource for Batch processing (opens in a new tab)
Concurrent requests
Authentication guide default: 3 in-flight requests/connections per model; organization-specific limits may differSource for Concurrent requests (opens in a new tab)

Pricing lists 5 for some detectors. Confirm the applicable organization quota; English diarized streaming may have a lower ceiling than plain transcription.

Maximum audio file
100 MB per prerecorded uploadSource for Maximum audio file (opens in a new tab)
API request fields
upload_fileSource for API request fields (opens in a new tab)

Only fields present in this endpoint’s reviewed request schema. A provider marketing option is not automatically a valid form field.

Audio formats
AAC, AIFF, FLAC, MP3, MP4, MOV, OGG, Opus, WAV, WebMSource for Audio formats (opens in a new tab)

Container acceptance implies processing its audio track, not video understanding.

Output formats
application/jsonSource for Output formats (opens in a new tab)

Prices retain their published currencies and billing units below.

Published billing units · original currency · observed 13 Oct 2026
CategoryUsageRateUnit
Audio analysisBatch Audio Event Detection1$0.02audio hour

Rate conditions

  1. Published self-service hourly audio rate, credit-based billing; taxes and negotiated enterprise terms are separate. Optional feature surcharges, where applicable, require the provider’s current account conditions. No text-token rate or cost per question is inferred.
Pricing source (opens in a new tab)

Data & enterprise terms

Terms: standard API content and outputs are not retained unless storage is enabled or required by the feature; stored content defaults to 35 days, with separately retained usage metadata and temporary residual backup/log copies. Self-service accounts are automatically enrolled in model improvement. FAQ narrows training to audio marked for storage; enterprise participation and opt-out depend on the agreement. Accent documents a per-request training_permitted=false opt-out. Do not infer a universal no-training or immediate-deletion guarantee.

Read the service policy (opens in a new tab)

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

Modulate Models API · batch

/api/velma-2-audio-event-classifier
Not verified

No discounted prompt-cache or persistent model-cache contract verified for these audio endpoints. Streaming buffers, stored clips and per-session speaker labels are not prompt caching.

How it works
Not collected

No discounted prompt-cache or persistent model-cache contract verified for these audio endpoints. Streaming buffers, stored clips and per-session speaker labels are not prompt caching.

Minimum reusable input
Not collected

No discounted prompt-cache or persistent model-cache contract verified for these audio endpoints. Streaming buffers, stored clips and per-session speaker labels are not prompt caching.

Lifetime & refresh
Not collected

No discounted prompt-cache or persistent model-cache contract verified for these audio endpoints. Streaming buffers, stored clips and per-session speaker labels are not prompt caching.

How to enable it
Not collected

No discounted prompt-cache or persistent model-cache contract verified for these audio endpoints. Streaming buffers, stored clips and per-session speaker labels are not prompt caching.

Usage fields
Not collected

No discounted prompt-cache or persistent model-cache contract verified for these audio endpoints. Streaming buffers, stored clips and per-session speaker labels are not prompt caching.

Billing rules
Not collected

No discounted prompt-cache or persistent model-cache contract verified for these audio endpoints. Streaming buffers, stored clips and per-session speaker labels are not prompt caching.

Limits & data boundaries
Not collected

No discounted prompt-cache or persistent model-cache contract verified for these audio endpoints. Streaming buffers, stored clips and per-session speaker labels are not prompt caching.

No verified cache rate for this endpoint. An empty rate is not a zero charge.

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Modulate presents these as its Velma model suite and audio-native proprietary technology. Its terms retain rights in algorithms and models and license hosted inference, with no public weight artifact identified. Entries represent the named closed managed model families, including ensembles, not invented individual internal checkpoints.

Classification evidence (opens in a new tab)

Sources reviewed