AI Models We Use

1. Writing, documents and slide decks

• Claude Fable 5.1 (Anthropic) — reports, business plans, meeting notes and outlines; the document editor’s drafting, rewriting and questions; slide decks; scripts and copy inside use cases; the translation step of video dubbing; and the AI assistants, including lyric writing.
• Qwen3.8 Flash (Alibaba Cloud) — rewriting, translation, summaries and product descriptions, and turning a music style you describe into tags.
• Qwen3.8 Omni Flash (Alibaba Cloud) — reads the files, recordings and videos you add to a document as sources.

2. Images

• GPT Image 2 (OpenAI) — image generation and editing: text to image, image editing, background, object, text and watermark removal, product and portrait photos, photo restoration, logos, stickers, thumbnails and the other image use cases.
• Tongyi Wanxiang (Alibaba Cloud) — creative posters, portrait style redraws, extending a photo to a new aspect ratio, and virtual-model and shoe-model product photos.
• Clarity Upscaler (open source) — image upscaling and sharpening. NAFNet (open source) — removes blur from photos.
• BiRefNet (open source) — cuts out the person for ID photos. InsightFace (open source) — finds the faces to blur in photos and videos.

3. Video

• HappyHorse 1.0 (Alibaba Cloud) — text to video, image to video, music videos, and video editing such as removing objects, subtitles, watermarks or backgrounds.
• Veo 3.1 (Google) — extends a video.
• Wan 2.2 Animate (Alibaba Cloud) — makes a still photo follow the motion of a reference video.
• SeedVR2 (ByteDance, open source) — video upscaling, enhancement and deblurring.
• Sync Lipsync (Sync Labs) — matches mouth movement in a video to audio.

4. Voice and transcription

• Qwen3 TTS Flash (Alibaba Cloud) — text to speech with preset voices, for read-aloud, audiobooks, voiceovers, dubbing and talking photos.
• Qwen3-TTS voice cloning and voice design (Alibaba Cloud) — creates a voice from your recording or from a description, and speaks with it.
• Qwen3.8 Omni Flash (Alibaba Cloud) — transcribes audio and video into timestamped, speaker-labelled text, including the original script for dubbing.

5. Music

• Suno V6 (Suno) — songs with vocals, and instrumentals.
• DiffRhythm (open source) — long-form songs.
• ACE-Step Audio-to-Audio (open-source model by ACE Studio and StepFun) — re-performs a song in a new style.
• Demucs HTDemucs (open-source model from Meta AI Research) — separates vocals, drums, bass and other stems, and removes vocals.
• Basic Pitch (open-source model from Spotify) — converts audio to MIDI.
• Audiveris and homr (open-source optical music recognition) — read sheet music from photos and PDFs.
• Mixing, trimming, slowed + reverb and joining video clips are ordinary audio and video processing (FFmpeg), not AI models.

6. Models you choose yourself

Our model catalog (wisaven.com/models) lists every model you can also run directly, each with its developer, parameters and price. It includes models from OpenAI, Google, Anthropic, Alibaba Cloud, ByteDance, Kuaishou, MiniMax, xAI, Microsoft, Meta, Black Forest Labs, DeepSeek, Moonshot AI, Zhipu AI, Tencent, Suno, ElevenLabs and Stability AI, among others. When you run a model from the catalog, that model makes your output.

7. Content safety and labelling

The same safety checks apply whichever model handles a request: the text you write is screened before generation (our blocklist and Waffo Prompt Sift), generated images and video frames are checked before they are delivered, and anything published to the gallery is moderated. See the Acceptable Use Policy (wisaven.com/legal/acceptable-use).

Everything these models produce is AI-generated. It may be inaccurate or resemble existing works; review it before you rely on it.