Audio/Video API Usage Notes
This document has details about what's currently supported, limitations, and workarounds for the Audio/Video APIs to help developers optimize their API implementations and understand service boundaries.
Text to Speech API usage
Here's the technical usage information for the Text to Speech API.
Limitations and workarounds
- TTS voice modulation: The output may have signification modulation in pitch or voice. Regenerating the audio can often resolve this issue.
- Limited voice controls: Currently voice controls like emphasis, speed or pitch modulation aren't supported.
- Mispronunciation: The audio output might mispronounce certain uncommon words or proper nouns. This can be addressed by using phonetic spellings.
Text input specifications
Transcript length: Up to 20000 characters.
Input format: Plain text, or .txt file via a pre-signed URL.
API render time
Render times for Text to Speech are 2X the output audio length.
Request limits
To be sure everyone enjoys peak performance with these APIs, Adobe sets limits on the volume, frequency, and concurrency of API calls. Adobe monitors your API usage and will contact you proactively to resolve any risks to API performance.
Be aware that these usage limits apply to your entire organization.
These are the current rate limits for API requests:
Get Voices API (/voices): 50 requests per minute.
Generate Speech API (/generate-speech): 10 requests per minute.
You may encounter a HTTP 429 "Too Many Requests" error if usage exceeds either the per minute or per day limits. We recommend using the retry-after header to determine the number of seconds to wait before trying again.
Reframe API usage
Supported media properties
| Attribute | Input | Output |
|---|---|---|
Formats | Video: .mp4, .mov; Image: .png, .gif | .mp4 |
Upload/Download type | Pre-signed URLs to individual videos, overlays | Pre-signed URLs to individual videos |
Video Duration (Max) | 30 minutes | Same as source |
Video Size (Max) | 10 GB | Same as source |
Video Codecs | H.265/HEVC (only 4:2:0), H.264/AVC | Same as source |
Color Properties | BT 601, BT 709, BT 2020, BT 2020 HLG, BT 2020 PQ | Same as source |
Frame Rate | 24, 25, 29.97, 30, 50, 59.94, 60 | Same as source |
4K Support | Yes | Yes |
Performance characteristics
Be aware that these characteristics apply when no focal point objects are specified in the payload of Reframe v2.
Estimated render times
| Aspect Ratios | Input Video Length | Scene Edit Detection | Estimated Render Time |
|---|---|---|---|
1 | 60s | No | ~0.5× video length |
5 | 60s | No | ~0.6× video length |
1 | 60s | Yes | ~1.3× video length |
5 | 60s | Yes | ~1.5× video length |
Reframing tips
When you're evaluating the suitability of your video for reframing, consider the following tips. For the best results, content should be in the FLY ZONE.
| 💚 FLY ZONE | ⚠️ NO FLY ZONE |
|---|---|
With Source Video | With Source Video |
- Clean footage (no graphics) - Multi-scene clips with trackable subjects that remain in the scene - Single-scene clips | - Graphics are embedded in the video - Video has multiple faces that need to be tracked - Content has letterboxing or pillarboxing applied |
- The number of keywords/phrases follow the guidelines - Brand names mentioned are on visible packaging - Multiple subjects are present. The system selects the largest frame area with the multiple keyword subjects in it | - Small objects in the scene or clip (like a football or baseball in sport footage) - Cannot set manual priority or weights for keywords/phrases - Negative keywords are used (like "exclude label", "avoid hands") - Positional words are used (like "leftmost", "center") - Celebrities or public figures are in keywords - Specialized terminology or jargon is used - Semantic nearness occurs. Common synonyms can overlap (e.g., bottle ≈ flask) - Long prose in the keyword. This is not a prompt |
Editability | Editability |
- Last mile editability is done in Premiere Pro 25.6.0+ - Scene Edit Detection is on/off - Letterbox is off | - Opening the .otio file in editors/versions other than Premiere Pro 25.6.0+ |
Request limits
To ensure equitable peak performance, Adobe limits the volume, frequency, and concurrency of API calls. We monitor usage to proactively resolve any risks to performance.
These are the current rate limits for API requests:
Reframe Processing API (/reframe): Max of 2 requests per minute.
You'll encounter a HTTP 429 - Too Many Requests error if usage exceeds the limits per minute or per day.
Use the retry-after header to determine the number of seconds you should wait before trying again.
Translate and Lip Sync API usage
Known limitations and workarounds
- Speaker Mismatch: Speaker mismatches or additional/missing speakers may occasionally occur in output transcripts. This has been observed in approximately 9% of cases. Content where speakers overlap may not produce the best results and should be avoided.
- Voice Modulation: Voices in the output may vary in pitch or show significant modulation. Regenerating the video/audio can often resolve this issue.
- Re-dubbing Dubbed Content: Avoid using deepfake content for re-dubbing purposes.
- Singing Isn't Supported: A music video or a song won't be dubbed correctly.
For editing transcripts
Only sentence editing is currently supported. Don't modify the timestamps.
Speakers can be updated, however don't remove speakers before dubbing. Also, dub using the edited transcripts in different target languages.
Language support
Dubbing is supported for the following languages:
| Language description | Code |
|---|---|
English (Indian) | en-IN |
English (American) | en-US |
English (British) | en-GB |
Spanish (Spanish) | es-ES |
Spanish (Argentina) | es-AR |
Spanish (Latin America) | es-419 |
French (France) | fr-FR |
French (Canada) | fr-CA |
Danish (Denmark) | da-DK |
Norwegian (Norway) | nb-NO |
German | de-DE |
Italian | it-IT |
Portuguese (Brazil) | pt-BR |
Portuguese (Portugal) | pt-PT |
Hindi (India) | hi-IN |
Japanese (Japan) | ja-JP |
Korean (South Korea) | ko-KR |
Input video support
Technical details for videos used as input:
- Duration (max): 30 mins
- FPS: 24 fps, 25 fps, 29.97, 30, 50, 59.94, 60
- Resolution (max): Full HD
1920*1080pxor1080*1920px - CODEC:
H.264, HEVC - Formats/container: .mp4, .mov
- Input medium: Pre-signed URL
- Render time: 3x the video length, 10x the video length (for 30 fps and 1080 resolution) if
lipSyncis enabled - Speaker speech (min): 5 secs
- Dubbing and Lip Sync: Multi-speaker support
Input audio support
Technical details for audio used as input:
- Duration (max): 30 mins
- CODEC:
MPEG, PCM - Formats/container: .mp3, .wav, .aac
- Input medium: Pre-signed URL
- Render time: 3x the audio length
- Dubbing: Multi-speaker support
Request limits
To ensure equitable peak performance, Adobe places limits on the volume, frequency, and concurrency of API calls, and monitors API usage to proactively resolve any risks to performance.
These usage limits apply to your entire organization.
The current limitations are:
Transcribe endpoint (/transcribe): 5 requests per minute.
Dubbing/Lip Sync endpoint (/dub): 5 requests per minute and 150 requests per day.
Avatar API usage
Known limitations and workarounds
- Gesture mismatch: Output videos may occasionally feature gesture mismatches.
Language support
Video generation is supported for the following languages:
| Language description | Code |
|---|---|
English (Indian) | en-IN |
English (American) | en-US |
English (British) | en-GB |
Spanish (Spanish) | es-ES |
Spanish (Argentina) | es-AR |
Spanish (Latin America) | es-419 |
French (France) | fr-FR |
French (Canada) | fr-CA |
Danish (Denmark) | da-DK |
Norwegian (Norway) | nb-NO |
German | de-DE |
Italian | it-IT |
Portuguese (Brazil) | pt-BR |
Portuguese (Portugal) | pt-PT |
Hindi (India) | hi-IN |
Japanese (Japan) | ja-JP |
Korean (South Korea) | ko-KR |
Avatar input audio specifications
Duration (max): 30 mins.
CODEC: MPEG, PCM.
Formats/container: audio/wav, audio/x-wav, audio/aac.
Input Medium: Pre-signed URL.
Avatar background video specifications
Duration (max): 30 mins.
FPS: 24 fps, 25 fps, 29.97, 30, 50, 59.94, 60.
Resolution (max): Full HD.
Aspect Ratio: 1,920*1,080px.
CODEC: H.264.
Formats/container: video/mp4, video/mov.
Input Medium: Pre-signed URL.
Avatar background image specifications
Formats: JPEG,PNG.
Input Medium: Pre-signed URL.
Aspect Ratio: 1,920*1,080px.
API render time
10X the output video length for Avatar API.
Request limits per API
To ensure equitable peak performance, Adobe places limits on the volume, frequency, and concurrency of API calls, and monitors API usage to proactively resolve any risks to performance.
These usage limits apply to your entire organization.
The current limitations are:
Get Actors API: 50 requests per minute.
Avatar API: 5 requests per minute.
GET status API
Request limits
To ensure equitable peak performance, Adobe places limits on the volume, frequency, and concurrency of API calls, and monitors API usage to proactively resolve any risks to performance.
These usage limits apply to your entire organization.
The current limitations are:
Get Result endpoint (/status/{jobId}): 100 requests per minute.
