PlansPlans that scale with you and your needs.
Starter$0/month + usage
Pro$249/month + usage
Usage Pricing
SolutionsApps readily available on Sieve that solve popular use cases out-of-the-box. Battle tested in production for best-in-class cost, quality, and speed.
| Name | Tag | Price |
|---|---|---|
| Dubbing (ElevenLabs voice) | sieve/dubbing | |
| Dubbing (OpenAI voice) | sieve/dubbing | |
| Lipsync | sieve/lipsync | |
| Autocrop | sieve/autocrop | |
| Active Speaker Detection | sieve/active_speaker_detection | |
| Speech Transcription | sieve/speech_transcriber | |
| Audio Enhance | sieve/audio_enhancement |
Optimized ModelsOpen-source video/audio models optimized by Sieve's team for best-in-class performance.
Visual Generative Models
| Name | Tag | Price |
|---|---|---|
| MuseTalk | sieve/musetalk | |
| CodeFormer | sieve/codeformer |
Visual Understanding Models
| Name | Tag | Price |
|---|---|---|
| SAM 2 (Segment Anything Model 2) Large | sieve/sam2 | |
| SAM 2 (Segment Anything Model 2) Tiny | sieve/sam2 | |
| TalkNet-ASD | sieve/talknet-asd |
Other public appsSieve also hosts a wide variety of other public apps that you can use to build your applications. These are priced by second of compute used by the request.
Sometimes, these apps call out to third-party model providers as backends for portions of the app (similar to dubbing). You can either enter your API key in Sieve's secret manager and be billed solely for the compute, or you can use those services through Sieve at-cost. Refer to the README sections of apps to understand when third-party integrations might be used.
| Compute | GPUs | vCPUs | RAM (GB) | Cost |
|---|---|---|---|---|
| A100 40GB | 1 | 12 | 85 | $4.20 / hr |
| A100 20GB | 1 | 6 | 42 | $2.80 / hr |
| L4 24GB | 1 | 8 | 32 | $1.25 / hr |
| T4 16GB | 1 | 4 | 16 | $0.81 / hr |
| CPU | 0 | 1 | 4 | $0.40 / hr |
Custom appsYou can also deploy custom functions running arbitrary code that are hosted and run on Sieve's infrastructure. This is useful for deploying custom models that aren't readily available on Sieve. These are billed based on compute, similar to public functions although you also pay the cost of loading models onto the GPU and any idle time these functions incur between requests.
Frequently asked questions