73 vendors

Ships image generation through the Gemini API as native multimodal output (the 'Nano Banana' model line) and, historically, through the standalone Imagen models. Also available via Vertex AI

ElevenLabs is primarily a speech AI company that also ships Eleven Music (text-to-song with vocals and lyrics) and a text-to-sound-effects model. Both are available in the web app and via AP

Alibaba's Tongyi Lab publishes the Wan family of video generation models. Wan 2.1 and Wan 2.2 (plus Wan-Animate) are released as open weights under Apache-2.0 on GitHub, Hugging Face and Mod

Publisher of the Stable Diffusion family plus a hosted developer platform (Stable Image services). Open weights ship under the Stability Community License, which is free below a $1M annual-r

Company founded by researchers from the original latent-diffusion work. Ships the FLUX family as both a hosted API (pro/flex/max) and downloadable weights (dev, klein) under a mix of Apache

Kuaishou Technology, a Beijing-headquartered short-video company listed in Hong Kong, builds and hosts the Kling AI generative video and image service. Kling is sold both as a consumer web/a

Provides image generation and editing through the gpt-image model family on the Images API and inside ChatGPT. Billing is token-based rather than a flat per-image rate.

Generative media company best known for the Dream Machine video models; its Photon image models share the same API platform and reference-image controls, which makes them convenient inside L

Pika makes a consumer AI video app focused on short-form clips and effect-driven edits (Pikaffects, Pikadditions, Pikaswaps, Pikaframes). It also runs a developer platform at dev.pika.art th

Generative video research company. Its Aleph model line is an in-context video-to-video model used for editing tasks (object add/remove, relight, restyle, new camera angles) alongside its Ge

Shengshu Technology develops the Vidu video generation model family, offered as a consumer product at vidu.com and as an enterprise/developer API at platform.vidu.com with credit-based billi

Maker of AI image and video enhancement software (Topaz Video, Topaz Photo, Gigapixel, Bloom, Astra, Project Starlight) sold as desktop apps, cloud apps and an API. Adobe announced on 2026-0

MiniMax Group is a Shanghai-based AI company that listed on the Hong Kong Stock Exchange in January 2026. Its video generation line is branded Hailuo AI for consumers (hailuoai.video) and ex

Tencent's Hunyuan division publishes HunyuanVideo and its derivatives (I2V, Avatar, HunyuanVideo 1.5) as downloadable weights. All are released under the Tencent Hunyuan Community License, w

Creative Cloud vendor whose Premiere Pro and Firefly products carry AI editing features including Generative Extend, AI Object Mask, Enhance Speech and Media Intelligence. Announced acquisit

Publisher of LatentSync, an open-weights audio-conditioned latent diffusion lipsync model released under Apache 2.0.

Creative AI suite built around the Magnific upscaling and relighting models plus a large licensed stock library. Freepik acquired Magnific AI in 2024 and subsequently rebranded the whole pla

AI avatar video platform offering photo avatars, consented digital twins, studio avatars, video translation and a pay-as-you-go API.

Higgsfield operates a consumer/prosumer generative media suite (Cinematic Studio, Marketing Studio, Canvas, Viral Presets, an After Effects plugin, a CLI and an MCP server) plus an asynchron

Image generation company focused on in-image typography and design layout. Runs a web editor plus a per-image API, and as of Ideogram 4 also publishes downloadable weights.

Lightricks develops the LTX-Video and LTX-2 open-weights video models alongside its LTX Studio product. Weights, inference code and (for LTX-2) training code are published on GitHub and Hugg

Meta's AI research group publishes AudioCraft, a research library containing MusicGen (text-to-music), AudioGen (text-to-sound), EnCodec, MAGNeT, MusicGen Style, JASCO and AudioSeal. The cod

PixVerse is a video generation platform founded in 2023 with teams across Asia and the US, which says it serves over 100 million creators in 175 countries and reached unicorn status in a Mar

Design-oriented generation platform whose distinguishing feature is native vector (editable SVG) output alongside raster. Sells a subscription plus a straightforward per-image API priced in

Suno, Inc. builds a consumer text-to-song platform plus a browser DAW (Suno Studio). Founded by four former Kensho employees (Mikey Shulman, Georg Kucsko, Martin Camacho, Keenan Freyberg) an

Lipsync and video dubbing API company. Ships the lipsync-* model family and the newer sync-3 model via Studio (web) and a developer API with SDKs.

ACE Studio and StepFun jointly released ACE-Step, an Apache-2.0 open-source music generation foundation model combining diffusion with a lightweight linear transformer. It targets fast full-

AIVA is a long-running AI composition tool aimed at video, game and content creators. It generates instrumental music from style presets and influence uploads and exports MP3, MIDI and WAV,

AWS offers Amazon Nova Reel, its first-party video generation model, exclusively through Amazon Bedrock. Nova Reel is positioned for enterprise use inside existing AWS accounts (IAM, VPC, Cl

Beatoven.ai generates background music and sound effects for video, podcast and game creators using its 'maestro' model, which the company says is trained on over 3 million ethically sourced

Maker of DaVinci Resolve, a professional edit/grade/VFX/audio suite whose Neural Engine provides a large set of AI tools. Sold as a free version plus a one-time Studio licence.

Realtime voice model company known for the Sonic text-to-speech family and the Ink speech-to-text family, aimed primarily at low-latency voice agents.

TTS startup that shut down in January 2024. Its XTTS-v2 model and the coqui-ai/TTS toolkit remain on GitHub and Hugging Face and are community-maintained, but no commercial license can be pu

Enterprise AI voice and dubbing company serving media localisation, live broadcast dubbing and voice agents, with TPN and GDPR certified infrastructure.

Speech API company best known for Nova speech-to-text; also ships the Aura text-to-speech family and a Voice Agent API. Positioning is voice agents and call handling rather than media produc

Transcript-driven video and podcast editor with an AI co-editor (Underlord) plus audio and video repair tools, and an API/MCP surface for paying users.

Epidemic Sound licenses a catalogue of human-made music, sound effects and AI voiceovers to video creators, and owns all rights in that catalogue so it can grant a single direct licence. Its

TTS and voice cloning platform with the S-series models (S1, S2 Pro, S2.1 Pro), flat per-byte API pricing and a free production-grade tier without latency guarantees.

Studio-facing neural performance editing company. Products DeepEditor and TrueSync are used for in-vision ADR, performance transfer, censorship editing and visual dubbing on feature and epis

Genmo is a San Francisco AI video research company that released Mochi 1, a 10B-parameter open-weights text-to-video diffusion model, under Apache-2.0 in late 2024. Genmo also runs a hosted

HPC-AI Tech, the team behind Colossal-AI, maintains Open-Sora, a fully Apache-2.0 open video generation stack including model weights, training and inference code, and published training cos

Haiper was a London-based video generation startup founded in 2021 by Google DeepMind alumni Yishu Miao and Ziyu Wang, which emerged from stealth with a $13.8m round. It no longer operates a

Character/avatar video company whose platform has expanded into a unified inference API fronting multiple image, video and audio models.

Builds emotionally expressive speech models: Octave text-to-speech (LLM-based, with voice design from text prompts) and EVI, a speech-to-speech conversational interface.

AI video generation and music-video styling tool (Superstudio). The domain still resolves and serves an application shell as of 2026-08-14 and no shutdown notice was found, but the site is f

Creative tool that aggregates many third-party models and, since Krea 2, ships its own from-scratch foundation image model with downloadable weights. Sells compute-unit subscriptions rather

Low-latency TTS API with 5-second voice cloning and 31 languages, priced per 1,000 characters with unlimited voice clones on all tiers.

Loudly runs an AI music platform combining its own generation models (VEGA instrumental models, MANTA vocal model), an AI remixer, AI mastering, a stem splitter and distribution to streaming

Enterprise speech platform with 200+ prebuilt neural voices, HD and multi-talker voice tiers, custom neural voice, speech-to-text in 130+ locales, and a Voice Live realtime speech-to-speech

Independent research lab that runs a subscription image (and now video) generation service accessed through Discord and its own web app. No official public API is documented; access is via s

New York based generative media company behind the Captions app and the Mirage family of audio/video foundation models. captions.ai/mirage now 301-redirects to mirage.app, indicating the mod

Moonvalley builds Marey, a video model it markets as trained exclusively on licensed, high-resolution footage with no scraped content, aimed at film and advertising production. Its positioni

Mubert generates royalty-free AI music for content, apps, games and streams, sold both as a consumer 'Render' subscription and as a developer API with per-generation quotas. Mubert Inc. reta

Voiceover product for video and e-learning with a studio editor plus an API (Gen2 for quality, Falcon 2 for streaming), voice changer, translation and dubbing.

NVIDIA publishes the Cosmos world foundation models, aimed primarily at robotics, autonomous driving and physical-AI simulation rather than creative filmmaking. Weights are released on Huggi

AI clipping and repurposing tool that turns long-form video into short vertical clips with captions, reframing, virality scoring and social publishing.

Pyramid Flow is an academic open-source project (ICLR 2025) from researchers affiliated with Peking University and Kuaishou. It is one of the few video models released under a plain MIT lice

Localisation services group. Papercup, the standalone AI dubbing startup, no longer operates independently: papercup.com 301-redirects to RWS's AI dubbing and voiceover service page with a '

Video localisation platform for translation, voice cloning and lip-synced dubbing, sold on minute-based plans with API access on higher tiers.

Originally a voice cloning and TTS company; its site and pricing now lead with deepfake detection, watermarking and identity verification products. Voice products (Resemble TTS, STS, Chatter

Speech-to-speech voice conversion and TTS company focused on film, TV and games, with a licensed voice marketplace and per-minute or per-character credit pricing.

Rhymes AI released Allegro, a small (2.8B) Apache-2.0 text-to-video model, in October 2024, along with the Aria multimodal LLM. Allegro was notable for fully permissive licensing at a time w

SOUNDRAW is a subscription AI music generator aimed at video creators and artists. It states its models are trained only on music written and recorded by its own in-house producers rather th

Skywork AI (associated with Kunlun Tech) publishes the SkyReels video models, notably SkyReels-V2 with a diffusion-forcing architecture for autoregressive long-video generation. Weights ship

Enterprise AI avatar video platform for training, comms and marketing video, built on its own Express avatar and voice engines with an API on higher tiers.

Udio, operated by Uncharted Labs, Inc., launched in April 2024 as a text-to-song generator. Founded December 2023 by David Ding, Conor Durkan, Charlie Nash, Yaroslav Ganin and Andrew Sanchez

Zhipu AI, spun out of Tsinghua University's Knowledge Engineering Group (THUDM), publishes the CogVideoX family. The 2B model is Apache-2.0; the 5B and CogVideoX1.5 models use a custom CogVi

Independent maintainer of Kokoro, an 82M-parameter open-weights TTS model released under Apache-2.0 with multiple language packs and dozens of voicepacks.

Independent researcher behind ControlNet, Fooocus and FramePack. FramePack is a next-frame-prediction wrapper that makes long video generation feasible on 6GB consumer GPUs; the code is Apac

Voice AI startup behind Play 3.0 mini and PlayDialog. Reported acquired by Meta in July 2025; secondary reporting says the commercial platform and API were wound down and the service termina

TTS provider targeting voice agents with two model lines: Coda for expressiveness and Mist v3 for very low latency. Publishes per-model TTFA figures and voice/language counts.

Voice company that open-sourced CSM-1B, a 1B-parameter conversational speech model, under Apache-2.0. Not included as a model entry because vendor-published specs could not be verified direc

Consumer text-to-speech reader that also sells a TTS API built on its Simba model family, with voice cloning from consented reference clips and SSML prosody controls.