Skip to main content
This section contains a description of rate limits and file upload specifications. deAPI is engineered for High Throughput capability upon deployment. Unlike traditional providers, API scalability is not restricted by complex access levels during the initial integration phase.

​
Account Types

All new accounts start as Basic and can upgrade to Premium by making any payment.
Upon registration, you receive a $5 bonus added to your balance. This bonus is available for Basic accounts with conservative rate limits designed for testing and evaluation.To upgrade to Premium with the model-specific limits below, simply top up your account with any available amount. Your remaining bonus balance carries over — for example, if you use $2 of your bonus and then purchase $10, your total balance becomes $13 ($3 remaining + $10 purchased), now usable with Premium limits.
Accounts registered with temporary or disposable email addresses do not receive the $5 bonus. We value long-term users and honest usage.

​
Rate Limits

Rate limits depend on the model, endpoint, and account tier. They are expressed as:
  • RPM — Requests Per Minute
  • RPD — Requests Per Day
Each window opens on your first request into a bucket and runs its full length from that moment — not to a calendar boundary in any timezone. Read X-RateLimit-Reset (RPM) and X-RateLimit-Daily-Reset (RPD) for the exact epoch second. The response headers describe the effective bucket. X-RateLimit-Source: model identifies a model override shared across account tiers; tier identifies an account-tier limit. A w=86400 entry in X-RateLimit-Policy indicates a daily limit. A minute-only policy means no daily limit applies; missing daily headers alone do not establish this.

​
Basic Limits

Basic accounts have conservative limits designed for testing and evaluation. For entries marked “Read response headers”, a fixed Basic quota is not published here. Read X-RateLimit-Limit for RPM and X-RateLimit-Policy for the applicable windows on a request made with your Basic account; do not assume the Premium quota applies.

​
Generation & Transformation Endpoints

In v2 audio and video transcription use a single endpoint (/api/v2/audio/transcriptions). Check the limits for the selected model in the table above.

​
Utility Endpoints

​
Premium Limits

Premium accounts use the model-specific RPM and RPD limits below. Utility endpoints have separate limits.
Transcription limits depend on the selected model: WhisperLargeV3 has 35 RPM / 11,000 RPD, and WhisperLargeV3Ct2 has 16 RPM / 4,600 RPD. Both use /api/v2/audio/transcriptions.
Video-description processing time depends on the source duration and preset. Read the rate-limit response headers for the limits effective on your account.

​
Utility Endpoints

Utility endpoints are limited generously rather than not at all, and they sit in their own buckets:
  • /api/v2/account/balance — 600 RPM (X-RateLimit-Source: static, identical for every account)
  • /api/v2/models — 300 RPM on Premium (X-RateLimit-Source: tier)
Price calculation and job status use the following tier limits: Read X-RateLimit-Limit on any response rather than hard-coding these.

​
Basic vs Premium Comparison

​
deAPI vs. Traditional Providers

Many AI infrastructure providers enforce restrictive initial limits, necessitating complex queue management systems for developers. Getting Started
  • Traditional Providers: Often require credit card verification or initial payment before any API access.
  • deAPI: Immediate access with $5 bonus upon registration — no payment required to start testing.
Scaling to Production
  • Traditional Providers: Typically require specific monetary spend thresholds (e.g., $50+) and time delays (weeks/months) to unlock higher rate limits.
  • deAPI: Instant upgrade to Premium with the model-specific limits above after any available top-up amount ($10, $25, or $50).
Limit Structure
  • Traditional Providers: Often utilize separate quotas for tokens (TPM), requests (RPM), daily limits (RPD), and per-model restrictions.
  • deAPI Premium: RPM and RPD quotas are listed per model, with separate limits for utility endpoints.
For High Volume Production requirements beyond Premium limits, please contact support via: support@deapi.ai

​
File Uploads

The majority of endpoints (Video-to-Text, Audio-to-Text, Image-to-Image) support direct binary file uploads or URL-based inputs. The global limits for file uploads are specified below.

​
Global Request & Response Cap

These caps apply to every endpoint regardless of media type. Exceeding them returns 413 Payload Too Large.
Per-type caps below are always ≤ the global 75 MB request body cap. If you send multiple files in one request, the sum must still fit within 75 MB.

​
Per-Type Upload Limits

1. Video (Video-to-Text, Image-to-Video, Video Upscale, Video Replace, Video Background Removal)
  • Max size: 50 MB
  • Supported Formats: MP4, MPEG, QuickTime (MOV), AVI, WMV, OGG
2. Audio (Audio-to-Text, Voice Cloning reference audio, Audio-to-Video, Music reference audio)
  • Max size: 20 MB
  • Supported Formats: AAC, MP3, OGG WebM
  • Voice Cloning reference audio: max 15 MB, 3–10 seconds duration. Supported formats: MP3, OGG, M4A.
  • Audio-to-Video conditioning audio: max 20 MB. Supported formats: MP3, OGG.
  • Music reference audio (style transfer): max 15 MB. Supported formats: MP3, OGG, M4A.
  • Image-to-Video reference audio (ref_audios): max 15 MB per clip. Supported formats: WAV, MP3, FLAC, OGG, M4A. Duration is model-specific (default range 1–10 seconds); audio must accompany at least one reference image or video.
3. Images (Image-to-Image, OCR, Background Removal, Upscale, Image-to-Video first/last frame, Reference Character)
  • Max size: 10 MB
  • Supported Formats: JPG, JPEG, PNG, GIF, BMP, WebP
  • Dimensions: min 64 px, max 5000 px per side
Image-to-Video references
  • Image-to-Video keyframe and reference images must be at least 1 KB per file.
  • ref_images uses the same image size, format, and dimension limits above, with an aspect ratio within 1:4–4:1. First/last keyframes may have any aspect ratio because they are fitted to the generation canvas.
  • Reference counts and durations depend on the selected model. Read info.limits.max_ref_images, max_ref_videos, max_ref_audios, and max_ref_files, plus min_ref_video_duration, max_ref_video_duration, min_ref_audio_duration, max_ref_audio_duration, min_ref_image_ratio, and max_ref_image_ratio. An unsupported reference type has a count limit of 0.
  • Outer ceilings are 9 images, 1 video, 3 audio clips, and 12 reference files in total. Reference video duration defaults to 2–5 seconds, but the model’s published limits take precedence. The global 75 MB request cap still applies.
  • References cannot be combined with first_frame_image or last_frame_image. See Video Animation for input modes and ordering rules.
For larger source media that won’t fit under the request body cap, prefer URL-based input (source_url) where supported — it bypasses the upload cap entirely and is subject only to the duration limits listed below.

​
Model Parameter Limits

Certain parameters — such as maximum character count, image resolution, frame count, or audio duration — are model-specific and may differ between models available on the same endpoint. These limits are returned dynamically via the Models endpoint:

​
URL Input & Duration Limits

For endpoints processing content via direct URLs (e.g., X Spaces), specific duration limits apply regardless of file size. 1. Audio-to-Text (X/Twitter Spaces)
  • Max Duration: 600 minutes
  • Note: Processing speeds for X Spaces may vary due to external bandwidth limits.
2. Video-to-Text (X, Kick, Twitch)
  • Max Duration: 600 minutes

​
Rate Limit Errors

When you exceed your rate limit, the API returns a 429 Too Many Requests response. How to handle:
  1. RPM exceeded — Wait 60 seconds before retrying
  2. RPD exceeded — Wait for the delay in Retry-After, or the epoch second in X-RateLimit-Daily-Reset
  3. Need higher limits? — Compare the Basic and Premium quotas for your model above. Upgrade to Premium with any payment, or contact support if your workload exceeds its limits.