AI Models We Use
The exact image, video, music and language models behind ohmymv, and how we use them.
Last updated: 2026-11-12
ohmymv is a production studio built on top of third-party AI models. We list them here in full so you know what your credits are buying. We are not affiliated with, endorsed by, or an official product of any of these model providers.
Image generation
| Model | Used for | | --------------------------------------------------------- | ---------------------------------------------------------------------- | | GPT Image 2.6 (OpenAI, via Apimart; Flare / Sunburst) | Storyboard keyframes, character reference frames, 2K turnaround sheets | | Google Gemini (image analysis) | Music segmentation, shot planning and automated QC of generated frames |
Video generation
| Model | Used for | | ---------------------------------------------------- | ---------------------------------------------------------------- | | MiniMax Hailuo H3 (via RunningHub) | First-frame image-to-video for cinematic B-roll shots | | MiniMax Hailuo H3 digital human (via RunningHub) | Singing A-roll clips with lip-sync, driven by your audio segment |
Music generation
| Model | Used for | | ---------------------------------- | ----------------------------------------------------- | | Suno (chirp v3.5) (via 302.AI) | Instrumental and vocal generation in the Music Studio |
Direction and analysis
| Model | Used for | | ----------------------------------- | ---------------------------------------------------------------------- | | Claude (Anthropic) (via 302.AI) | Director bible, shot planning, prompt writing and storyboard rewriting | | Google Gemini | Audio structure analysis, lyric alignment and video review |
Assembly
Final masters are assembled locally with FFmpeg: EBU R128 loudness normalisation (loudnorm, -16 LUFS / -1.5 dBTP), 0.25–0.35s optical crossfades with a 1.5s opening and 3.5s closing fade, and burned-in bilingual subtitles, rendered to a 1080p / 24fps canvas. Assembly is deterministic and does not use a generative model.
What this means for your output
- Every image, clip and track is AI-generated. It is not footage of a real event.
- Model output is stochastic: the same prompt can produce different results between runs.
- You can see which model produced a given asset in your asset library metadata.
- We do not fine-tune these providers' base models on your content without your explicit consent.
Content safety
Every prompt is screened by Waffo Prompt Sift before a model is called, and generated output is reviewed before delivery. See our Acceptable Use Policy for what may not be generated.
Questions
support@ohmymv.com