Search the full model catalog.

Google's Gemini 2.5 Flash reads text, images, video and audio and answers in text, with adjustable thinking, function calling and JSON schema output over a context of about one million tokens.

DreamOmni2 edits or composes images from a text instruction plus up to four input pictures. A reference can supply a subject, like a jacket or a pet, or an abstract trait such as a print or a painting style.

Lucy Edit Dev applies a written instruction to an existing video, such as swapping an outfit or turning a person into a creature, while the original motion and framing stay in place. Common edits need no masks.

Builds a full 360-degree equirectangular panorama around a single photo, generating the parts of the scene the camera never captured. You set the field of view the photo covers on the sphere.

Edits a photo from a plain-language instruction and can take up to two more images as references, for example a logo to print on a shirt. With no image attached, it creates one from the prompt.

The lighter variant in the HiDream-I1 family: the 17B Full model distilled to 16 reference sampling steps, run with guidance set to 0.

The undistilled 17B-parameter HiDream-I1 text-to-image checkpoint. It is the only one of the three built to run with classifier-free guidance, so the guidance scale and negative prompt both shape the image.

Reads an employee leave export, such as a CSV, and returns JSON with leave counts by type, month and department, the top 10 employees by leave days, and a written summary in English or Turkish.

Qwen's 20B-parameter MMDiT text-to-image model. It draws legible English and Chinese text into the picture, from a single shop sign to a multi-line poster.

Speech to text from a Turkish fine-tune of OpenAI Whisper large-v3-turbo. Diarization, on by default, puts a start and end time on each line, and setting 2 to 5 speakers adds speaker labels.

PixVerse V4.5 generates video from one photo and a text prompt, at 360p to 1080p and 5 or 8 seconds per clip. It is the newest of the V3.5, V4 and V4.5 image-to-video versions and shares their controls.

Repairs cracked or faded old photos and returns them in color. There is no prompt: you upload one image and the app redraws it.

Turns an everyday photo of a person into a business headshot: same face, professional clothing, and a white, black, gray, neutral or office background.

MiniMax's Hailuo S2V-01 makes a video of a person from one face photo and a prompt, keeping their facial features consistent while the prompt sets the scene.

Hailuo 01 (MiniMax I2V-01) turns one still image into a 6-second 720p video at 25 fps, using the image as the first frame and an optional prompt to direct motion and camera.

Turns a topic prompt into a narrated vertical short with captions synced to the voice and an opening hook title. Six video models are available to render the scenes.

Covers a YouTube song in a different voice. A ready-made or custom RVC voice model replaces the lead vocal, and the app mixes it back over the original music and backing vocals.

Puts one or two faces into a scene it generates from your prompt, or into a photo you upload. In a two-person photo it can replace just the left or the right person.

Turns your lyrics, written with [verse] and [chorus] markers, into a sung track shaped by genre, instrument, mood and voice tags, using the open ACE-Step v1 3.5B music model.

Chat with one branch of a GitHub repository: its files are indexed, and the closest matches go to the chat model you select. Each run needs a GitHub access token, public repositories included.