Montage media models (managed catalog)
When a Navin Plus+ plan (or equivalent managed key) is active, Studio → Montage → AI media providers lists curated models. You can also point each modality at BYOK providers under Settings.
Defaults below match the shipped managed catalog (managed_models / managed_catalog fallback). Unit prices are catalog estimates; billed amount follows provider usage.cost when present. Local FFmpeg packaging and HyperFrames encode do not bill media APIs.
Defaults (recommended)
| Modality | Default model | Role in the GTM loop |
|---|
| Image | google/gemini-3.1-flash-image (Nano Banana 2) | Product / lifestyle / social stills tied to the linked app |
| Video | google/veo-3.1-fast | Short AI B-roll / promo clips (default ~720p / 8 s estimate) |
| Music | google/lyria-3-clip-preview | 30 s bed / jingle (0,04 $ / clip) |
| TTS | google/gemini-3.1-flash-tts-preview | Voiceover for demos and exports |
| STT | qwen/qwen3-asr-flash-2026-02-10 | Captions / dictation from recorded audio |
Image
| Slug | Name | Notes |
|---|
google/gemini-3.1-flash-image | Nano Banana 2 | Default - quality/price balance |
google/gemini-3-pro-image | Nano Banana Pro | Premium 2K/4K |
bytedance-seed/seedream-4.5 | Seedream 4.5 | Economy ~0,04 $ / image |
Video
| Slug | Name | Notes |
|---|
google/veo-3.1-fast | Veo 3.1 Fast | Default + audio; 720p ≈ 0,10 $/s |
google/veo-3.1 | Veo 3.1 | Premium from ≈ 0,40 $/s |
bytedance/seedance-2.5 | Seedance 2.5 | References / edit workflows |
kwaivgi/kling-v3.0-pro | Kling v3.0 Pro | Cinematic control (multi-SKU) |
bytedance/seedance-2.0-fast | Seedance 2.0 Fast | Token-billed; 720p/8 s ≈ 0,97 $ |
Prefer a live browser demo of the web/mobile UI for authenticity; use AI video for B-roll, transitions, or when no runnable UI exists yet.
Music
| Slug | Name | Notes |
|---|
google/lyria-3-clip-preview | Lyria 3 Clip | Default - 30 s, 0,04 $ / clip |
google/lyria-3-pro-preview | Lyria 3 Pro | Full song on request - 0,08 $ / song |
TTS (voice out)
| Slug | Name | Notes |
|---|
google/gemini-3.1-flash-tts-preview | Gemini 3.1 Flash TTS | Default - 70+ languages |
fish-audio/s2.1-pro-free:free | Fish Audio free | Prototyping only |
x-ai/grok-voice-tts-1.0 | Grok Voice TTS | Multilingual voices |
fish-audio/s2.1-pro | Fish Audio Pro | Paid production fallback |
STT (voice in / captions)
| Slug | Name | Notes |
|---|
qwen/qwen3-asr-flash-2026-02-10 | Qwen3 ASR Flash | Default - ≈ 0,0021 $ / min |
openai/gpt-transcribe | GPT Transcribe | 0,0045 $ / min |
nvidia/parakeet-tdt-0.6b-v3 | Parakeet TDT 0.6B v3 | 0,0015 $ / min |
microsoft/mai-transcribe-1.5 | MAI-Transcribe 1.5 | ≈ 0,006 $ / min, 43 languages |
deepgram/nova-3 | Deepgram Nova-3 | from ≈ 0,0043 $ / min |
Multimodal text (vision route)
Used when the agent reasons over screenshots / frames (not the Montage picker itself):
| Slug | Name |
|---|
google/gemini-3.6-flash | Gemini 3.6 Flash (main multimodal) |
xiaomi/mimo-v2.5 | MiMo V2.5 (economy multimodal) |
Access
- Plus+ managed key - pick models in Montage or Settings.
- BYOK - configure providers, then the same modality settings.
- FFmpeg-only - package an existing demo with zero media API spend.
Never auto-publish. Hand off to /ads only with explicit approval.
