Search the full model catalog.

Multilingual text-to-speech from Hume AI that clones a voice from a reference clip, built on Llama 3.2 3B. Its tokenizer pairs each text token with one acoustic vector, which keeps the audio aligned to the script.

Generates high-quality speech from text using advanced TTS technology with support for voice cloning and multi-speaker synthesis.

FireRed Image Edit 1.1 is the March 2026 update of FireRedTeam's 20B image editor. It keeps a person's identity stable across edits and combines several reference images, such as a portrait and clothing, in one result.

Generates natural-sounding speech from text with support for multi-speaker voice cloning and fast inference capabilities.

A 3B OCR model from rednote-hilab that turns PDFs and images into Markdown with HTML tables and LaTeX formulas, plus a labeled box for every block. Its makers designed it to read almost any human script.

Qwen3.5-27B is a 27B model from Qwen's Qwen3.5 line that always reasons before it replies and can call your functions. Qwen reports 72.4 on SWE-bench Verified for it.

Speech-to-text model from Qwen's Qwen3-ASR family, built on Qwen3-Omni. It transcribes speech in 30 languages, and Qwen also tested it on singing and full songs with background music.

Generate high-quality images using Seedream V5 Lite, supporting both image-to-image and text-to-image transformations with customizable settings.

Mistral AI's 4B streaming speech-to-text model. It writes out microphone audio in 13 languages while the speaker is still talking, with a transcription delay you set.

Generate expressive speech in 23 languages with zero-shot voice cloning from a short reference clip, an emotion exaggeration control and cross-language voice transfer.

OpenMOSS's 8B dialogue model turns a speaker-tagged script into one audio track with up to five voices, each of which can be cloned from a 3 to 10 second clip.

Streaming text-to-speech from OpenMOSS that sends audio in chunks while the rest of the text is still being synthesized. Clones a voice from one reference clip and speaks 20 languages.

Turns one or two product photos into a short studio video: a front-to-back product turn, a ghost mannequin garment shot or one of four camera moves, in 9:16, 16:9 or 1:1.

Qwen's 1.7B-parameter text-to-speech model with nine preset voices across 10 languages. A short written instruction sets the emotion and pace of each line.

Google's TranslateGemma 4B applied to images: it reads the writing in a photo, screenshot or scan and returns it translated into the target language as plain text.

Google's largest TranslateGemma model, used here to read the text in a picture and return it in another language. Its card gives it the family's best scores on both text and image translation tests.

The largest model in Google's TranslateGemma family, a 27B Gemma 3 fine-tune that translates text across 55 languages. It has the best scores in the family on every benchmark in Google's model card.

The mid-sized model in Google's TranslateGemma family: Gemma 3 fine-tuned for text translation across 55 languages. Google reports it outperforms the larger Gemma 3 27B baseline on WMT24++.

Google's 4B-parameter open translation model, fine-tuned from Gemma 3 for text in 55 languages. Google reports it rivals the larger Gemma 3 12B baseline on WMT24++.

A compact AI model optimized for text generation tasks, designed for efficient processing and accurate responses.